bump version from 0.23.0 to 0.24.0rc1

101 AP algo: do not interpollate precision

.github/workflows/notebook-bot.yml

@@ -0,0 +1,64 @@
+name: Notebook Check Pull Request
+
+on:
+  pull_request_target:
+    types: [opened, reopened]
+
+permissions:
+  contents: read
+
+jobs:
+  comment-welcome:
+    permissions:
+      contents: read
+      pull-requests: write
+
+    runs-on: ubuntu-latest
+    steps:
+    - name: Fetch pull request branch
+      uses: actions/checkout@v4
+      with:
+        repository: ${{ github.event.pull_request.head.repo.full_name }}
+        ref: ${{ github.event.pull_request.head.sha }}
+    - name: Fetch base develop branch
+      run: git fetch -u "$GITHUB_SERVER_URL/$GITHUB_REPOSITORY" develop:develop
+    - name: Create message
+      env:
+        HEAD_REPOSITORY: ${{ github.event.pull_request.head.repo.full_name }}
+        HEAD_REF: ${{ github.event.pull_request.head.ref }}
+        PR_NUM: ${{ github.event.pull_request.number }}
+      run: |
+        # Preview links and tool usage only needed for notebook changes.
+        readarray -t changed_notebooks < <(git diff --name-only develop | grep '\.ipynb$' || true)
+        if [[ ${#changed_notebooks[@]} == 0 ]]; then
+          echo "No notebooks modified in this pull request."
+        else
+          msg="<h4>Preview</h4>\n"
+          msg+="Preview and run these notebook edits with Google Colab:\n<ul>\n"
+          # Link to PR branch in user's fork that is always current.
+          for fp in "${changed_notebooks[@]}"; do
+            gh_path="${HEAD_REPOSITORY}/blob/${HEAD_REF}/${fp}"
+            colab_url="https://colab.research.google.com/github/${gh_path}"
+            msg+="<li><a href='${colab_url}'>${fp}</a></li>\n"
+          done
+          msg+="</ul>\n"
+
+          reviewnb_url="https://app.reviewnb.com/${GITHUB_REPOSITORY}/pull/${PR_NUM}/files/"
+          msg+="Rendered <a href='${reviewnb_url}'>notebook diffs</a> available on ReviewNB.com.\n"
+
+          msg+="If commits are added to the pull request, synchronize your local branch: <code>git pull origin $HEAD_REF</code>\n"
+        fi
+        echo "MESSAGE=$msg" >> $GITHUB_ENV
+    - name: Post comment
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        ISSUE_URL: ${{ github.event.pull_request.issue_url }}
+      run: |
+        # Env var defined in previous step. Escape string for JSON.
+        body="$(echo -n -e $MESSAGE | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read()))')"
+        # Add comment to pull request.
+        curl -X POST \
+          -H "Accept: application/vnd.github.v3+json" \
+          -H "Authorization: token $GITHUB_TOKEN" \
+          "${ISSUE_URL}/comments" \
+          --data "{\"body\": $body}"

.github/workflows/test-doc.yml

@@ -14,21 +14,8 @@ jobs:
         uses: actions/setup-python@v5
         with:
           python-version: '3.10'
-      - name: 📦 Install mkdocs-material
-        run: pip install "mkdocs-material[all]"
-      - name: 📦 Install mkdocstrings[python]
-        run: pip install "mkdocstrings[python]"
-      - name: 📦 Install mkdocs-material[imaging]
-        run: pip install "mkdocs-material[imaging]"
-      - name: 📦 Install mike
-        run: pip install "mike"
-      - name: 📦 Install mkdocs-git-revision-date-localized-plugin
-        run: pip install "mkdocs-git-revision-date-localized-plugin"
-      - name: 📦 Install JupyterLab
-        run: pip install jupyterlab
-      - name: 📦 Install mkdocs-jupyter
-        run: pip install mkdocs-jupyter
-      - name: 📦 Install mkdocs-git-committers-plugin-2
-        run: pip install mkdocs-git-committers-plugin-2
-      - name: 🧪 Test documentation build
-        run: mkdocs build --verbose
+      - name:  🏗️ Install dependencies and Test Docs Build
+        run: |
+          python -m pip install --upgrade pip
+          pip install "mkdocs-material" "mkdocstrings[python]" "mkdocs-material[imaging]" mike "mkdocs-git-revision-date-localized-plugin" jupyterlab mkdocs-jupyter mkdocs-git-committers-plugin-2
+          mkdocs build --verbose

.pre-commit-config.yaml

@@ -9,7 +9,6 @@ repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
     rev: v4.6.0
     hooks:
-      - id: end-of-file-fixer
       - id: trailing-whitespace
         exclude: test/.*\.py
       - id: check-yaml
@@ -19,33 +18,21 @@ repos:
       - id: check-case-conflict
       - id: check-added-large-files
       - id: detect-private-key
-      - id: forbid-new-submodules
       - id: pretty-format-json
         exclude: demo.ipynb
         args: ['--autofix', '--no-sort-keys', '--indent=4']
       - id: end-of-file-fixer
       - id: mixed-line-ending
 
-
   -   repo: https://github.com/PyCQA/bandit
-      rev: '1.7.8'
+      rev: '1.7.9'
       hooks:
       -   id: bandit
           args: ["-c", "pyproject.toml"]
           additional_dependencies: ["bandit[toml]"]
 
-  -   repo: https://github.com/pycqa/isort
-      rev: 5.13.2
-      hooks:
-      -   id: isort
-          name: isort (python)
-      -   id: isort
-          name: isort (pyi)
-          types: [pyi]
-
-
   -   repo: https://github.com/astral-sh/ruff-pre-commit
-      rev: v0.4.1
+      rev: v0.6.3
       hooks:
       -   id: ruff
           args: [--fix, --exit-non-zero-on-fix]

CONTRIBUTING.md

@@ -4,6 +4,25 @@ Thank you for your interest in contributing to Supervision!
 
 We are actively improving this library to reduce the amount of work you need to do to solve common computer vision problems.
 
+## Code of Conduct
+
+Please read and adhere to our [Code of Conduct](CODE_OF_CONDUCT.md). This document outlines the expected behavior for all participants in our project.
+
+## Table of Contents
+
+- [Contribution Guidelines](#contribution-guidelines)
+  - [Contributing Features](#contributing-features-)
+- [How to Contribute Changes](#how-to-contribute-changes)
+- [Installation for Contributors](#installation-for-contributors)
+- [Code Style and Quality](#-code-style-and-quality)
+  - [Pre-commit tool](#pre-commit-tool)
+  - [Docstrings](#docstrings)
+  - [Type checking](#type-checking)
+- [Documentation](#-documentation)
+- [Cookbooks](#-cookbooks)
+- [Tests](#-tests)
+- [License](#-license)
+
 ## Contribution Guidelines
 
 We welcome contributions to:
@@ -32,49 +51,49 @@ First, fork this repository to your own GitHub account. Click "fork" in the top
 
 Then, run `git clone` to download the project code to your computer.
 
+You should also set up `roboflow/supervision` as an "upstream" remote (that is, tell git that the reference Supervision repository was the source of your fork of it):
+
+```bash
+git remote add upstream https://github.com/roboflow/supervision.git
+git fetch upstream
+```
+
 Move to a new branch using the `git checkout` command:
 
 ```bash
-git checkout -b <your_branch_name>
+git checkout -b <scope>/<your_branch_name> upstream/develop
 ```
 
-The name you choose for your branch should describe the change you want to make (i.e. `line-counter-docs`).
+The name you choose for your branch should describe the change you want to make and start with an appropriate prefix:
+
+- `feat/`: for new features (e.g., `feat/line-counter`)
+- `fix/`: for bug fixes (e.g., `fix/memory-leak`)
+- `docs/`: for documentation changes (e.g., `docs/update-readme`)
+- `chore/`: for routine tasks, maintenance, or tooling changes (e.g., `chore/update-dependencies`)
+- `test/`: for adding or modifying tests (e.g., `test/add-unit-tests`)
+- `refactor/`: for code refactoring (e.g., `refactor/simplify-algorithm`)
 
 Make any changes you want to the project code, then run the following commands to commit your changes:
 
 ```bash
-git add .
-git commit -m "Your commit message"
-git push -u origin main
+git add -A
+git commit -m "feat: add line counter functionality"
+git push -u origin <your_branch_name>
 ```
 
-## 🎨 Code quality
+Use conventional commit messages to clearly describe your changes. The format is:
 
-### Pre-commit tool
+<type>[optional scope]: <description>
 
-This project uses the [pre-commit](https://pre-commit.com/) tool to maintain code quality and consistency. Before submitting a pull request or making any commits, it is important to run the pre-commit tool to ensure that your changes meet the project's guidelines.
-
-Furthermore, we have integrated a pre-commit GitHub Action into our workflow. This means that with every pull request opened, the pre-commit checks will be automatically enforced, streamlining the code review process and ensuring that all contributions adhere to our quality standards.
-
-To run the pre-commit tool, follow these steps:
-
-1. Install pre-commit by running the following command: `poetry install`. It will not only install pre-commit but also install all the deps and dev-deps of project
-
-2. Once pre-commit is installed, navigate to the project's root directory.
-
-3. Run the command `pre-commit run --all-files`. This will execute the pre-commit hooks configured for this project against the modified files. If any issues are found, the pre-commit tool will provide feedback on how to resolve them. Make the necessary changes and re-run the pre-commit command until all issues are resolved.
-
-4. You can also install pre-commit as a git hook by execute `pre-commit install`. Every time you made `git commit` pre-commit run automatically for you.
-
-### Docstrings
-
-All new functions and classes in `supervision` should include docstrings. This is a prerequisite for any new functions and classes to be added to the library.
-
-`supervision` adheres to the [Google Python docstring style](https://google.github.io/styleguide/pyguide.html#383-functions-and-methods). Please refer to the style guide while writing docstrings for your contribution.
-
-### Type checking
-
-So far, **there is no type checking with mypy**. See [issue](https://github.com/roboflow-ai/template-python/issues/4).
+Common types include:
+- feat: A new feature
+- fix: A bug fix
+- docs: Documentation only changes
+- style: Changes that do not affect the meaning of the code (white-space, formatting, etc)
+- refactor: A code change that neither fixes a bug nor adds a feature
+- perf: A code change that improves performance
+- test: Adding missing tests or correcting existing tests
+- chore: Changes to the build process or auxiliary tools and libraries
 
 Then, go back to your fork of the `supervision` repository, click "Pull Requests", and click "New Pull Request".
 
@@ -104,7 +123,75 @@ All pull requests will be reviewed by the maintainers of the project. We will pr
 
 PRs must pass all tests and linting requirements before they can be merged.
 
-## 📝 documentation
+## Installation for Contributors
+
+Before starting your work on the project, set up your development environment:
+
+1. Clone your fork of the project:
+   ```bash
+   git clone https://github.com/YOUR_USERNAME/supervision.git
+   cd supervision
+   ```
+   Replace `YOUR_USERNAME` with your GitHub username.
+
+2. Create and activate a virtual environment:
+   ```bash
+   python3 -m venv .venv
+   source .venv/bin/activate
+   ```
+
+3. Install Poetry:
+
+   Using pip:
+   ```bash
+   pip install -U pip setuptools
+   pip install poetry
+   ```
+
+   Or using pipx (recommended for global installation):
+   ```bash
+   pipx install poetry
+   ```
+
+4. Install project dependencies:
+   ```bash
+   poetry install
+   ```
+
+5. Run pytest to verify the setup:
+   ```bash
+   poetry run pytest
+   ```
+
+## 🎨 Code Style and Quality
+
+### Pre-commit tool
+
+This project uses the [pre-commit](https://pre-commit.com/) tool to maintain code quality and consistency. Before submitting a pull request or making any commits, it is important to run the pre-commit tool to ensure that your changes meet the project's guidelines.
+
+Furthermore, we have integrated a pre-commit GitHub Action into our workflow. This means that with every pull request opened, the pre-commit checks will be automatically enforced, streamlining the code review process and ensuring that all contributions adhere to our quality standards.
+
+To run the pre-commit tool, follow these steps:
+
+1. Install pre-commit by running the following command: `poetry install --with dev`. It will not only install pre-commit but also install all the deps and dev-deps of project
+
+2. Once pre-commit is installed, navigate to the project's root directory.
+
+3. Run the command `pre-commit run --all-files`. This will execute the pre-commit hooks configured for this project against the modified files. If any issues are found, the pre-commit tool will provide feedback on how to resolve them. Make the necessary changes and re-run the pre-commit command until all issues are resolved.
+
+4. You can also install pre-commit as a git hook by executing `pre-commit install`. Every time you do a `git commit` pre-commit run automatically for you.
+
+### Docstrings
+
+All new functions and classes in `supervision` should include docstrings. This is a prerequisite for any new functions and classes to be added to the library.
+
+`supervision` adheres to the [Google Python docstring style](https://google.github.io/styleguide/pyguide.html#383-functions-and-methods). Please refer to the style guide while writing docstrings for your contribution.
+
+### Type checking
+
+So far, **there is no type checking with mypy**. See [issue](https://github.com/roboflow-ai/template-python/issues/4).
+
+## 📝 Documentation
 
 The `supervision` documentation is stored in a folder called `docs`. The project documentation is built using `mkdocs`.
 
@@ -112,13 +199,13 @@ To run the documentation, install the project requirements with `poetry install
 
 You can learn more about mkdocs on the [mkdocs website](https://www.mkdocs.org/).
 
-## 🧑‍🍳 cookbooks
+## 🧑‍🍳 Cookbooks
 
 We are always looking for new examples and cookbooks to add to the `supervision`
 documentation. If you have a use case that you think would be helpful to others, please
 submit a PR with your example. Here are some guidelines for submitting a new example:
 
-- Create a new notebook in the [`docs/nodebooks`](https://github.com/roboflow/supervision/tree/develop/docs/notebooks) folder.
+- Create a new notebook in the [`docs/notebooks`](https://github.com/roboflow/supervision/tree/develop/docs/notebooks) folder.
 - Add a link to the new notebook in [`docs/theme/cookbooks.html`](https://github.com/roboflow/supervision/blob/develop/docs/theme/cookbooks.html). Make sure to add the path to the new notebook, as well as a title, labels, author and supervision version.
 - Use the [Count Objects Crossing the Line](https://supervision.roboflow.com/develop/notebooks/count-objects-crossing-the-line/) example as a template for your new example.
 - Freeze the version of `supervision` you are using.
@@ -126,10 +213,10 @@ submit a PR with your example. Here are some guidelines for submitting a new exa
 - Notebook should be self-contained. If you rely on external data ( videos, images, etc.) or libraries, include download and installation commands in the notebook.
 - Annotate the code with appropriate comments, including links to the documentation describing each of the tools you have used.
 
-## 🧪 tests
+## 🧪 Tests
 
 [`pytests`](https://docs.pytest.org/en/7.1.x/) is used to run our tests.
 
-## 📄 license
+## 📄 License
 
 By contributing, you agree that your contributions will be licensed under an [MIT license](https://github.com/roboflow/supervision/blob/develop/LICENSE.md).

README.md

@@ -16,6 +16,7 @@
 
 [![version](https://badge.fury.io/py/supervision.svg)](https://badge.fury.io/py/supervision)
 [![downloads](https://img.shields.io/pypi/dm/supervision)](https://pypistats.org/packages/supervision)
+[![snyk](https://snyk.io/advisor/python/supervision/badge.svg)](https://snyk.io/advisor/python/supervision)
 [![license](https://img.shields.io/pypi/l/supervision)](https://github.com/roboflow/supervision/blob/main/LICENSE.md)
 [![python-version](https://img.shields.io/pypi/pyversions/supervision)](https://badge.fury.io/py/supervision)
 [![colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow/supervision/blob/main/demo.ipynb)
@@ -28,7 +29,7 @@
 
 **We write your reusable computer vision tools.** Whether you need to load your dataset from your hard drive, draw detections on an image or video, or count how many detections are in a zone. You can count on us! 🤝
 
-[![supervision-hackfest](https://github.com/roboflow/supervision/assets/26109316/c05cc954-b9a6-4ed5-9a52-d0b4b619ff65)](https://github.com/orgs/roboflow/projects/10)
+[![supervision-hackfest](https://github.com/roboflow/supervision/assets/26109316/c05cc954-b9a6-4ed5-9a52-d0b4b619ff65)](https://github.com/orgs/roboflow/projects)
 
 ## 💻 install
 
@@ -86,7 +87,7 @@ len(detections)
 
 ### annotators
 
-Supervision offers a wide range of highly customizable [annotators](https://supervision.roboflow.com/latest/annotators/), allowing you to compose the perfect visualization for your use case.
+Supervision offers a wide range of highly customizable [annotators](https://supervision.roboflow.com/latest/detection/annotators/), allowing you to compose the perfect visualization for your use case.
 
 ```python
 import cv2
@@ -95,8 +96,8 @@ import supervision as sv
 image = cv2.imread(...)
 detections = sv.Detections(...)
 
-bounding_box_annotator = sv.BoundingBoxAnnotator()
-annotated_frame = bounding_box_annotator.annotate(
+box_annotator = sv.BoxAnnotator()
+annotated_frame = box_annotator.annotate(
     scene=image.copy(),
     detections=detections
 )
@@ -106,22 +107,25 @@ https://github.com/roboflow/supervision/assets/26109316/691e219c-0565-4403-9218-
 
 ### datasets
 
-Supervision provides a set of [utils](https://supervision.roboflow.com/latest/datasets/) that allow you to load, split, merge, and save datasets in one of the supported formats.
+Supervision provides a set of [utils](https://supervision.roboflow.com/latest/datasets/core/) that allow you to load, split, merge, and save datasets in one of the supported formats.
 
 ```python
 import supervision as sv
+from roboflow import Roboflow
 
-dataset = sv.DetectionDataset.from_yolo(
-    images_directory_path=...,
-    annotations_directory_path=...,
-    data_yaml_path=...
+project = Roboflow().workspace(<WORKSPACE_ID>).project(<PROJECT_ID>)
+dataset = project.version(<PROJECT_VERSION>).download("coco")
+
+ds = sv.DetectionDataset.from_coco(
+    images_directory_path=f"{dataset.location}/train",
+    annotations_path=f"{dataset.location}/train/_annotations.coco.json",
 )
 
-dataset.classes
-['dog', 'person']
+path, image, annotation = ds[0]
+    # loads image on demand
 
-len(dataset)
-# 1000
+for path, image, annotation in ds:
+    # loads image on demand
 ```
 
 <details close>
@@ -216,7 +220,7 @@ len(dataset)
 
 ## 🎬 tutorials
 
-Want to learn how to use Supervision? Explore our [how-to guides](https://supervision.roboflow.com/develop/how_to/detect_and_annotate/), [end-to-end examples](https://github.com/roboflow/supervision/tree/develop/examples), and [cookbooks](https://supervision.roboflow.com/develop/cookbooks/)!
+Want to learn how to use Supervision? Explore our [how-to guides](https://supervision.roboflow.com/develop/how_to/detect_and_annotate/), [end-to-end examples](https://github.com/roboflow/supervision/tree/develop/examples), [cheatsheet](https://roboflow.github.io/cheatsheet-supervision/), and [cookbooks](https://supervision.roboflow.com/develop/cookbooks/)!
 
 <br/>
 
@@ -291,7 +295,7 @@ We love your input! Please see our [contributing guide](https://github.com/robof
           />
       </a>
       <img src="https://raw.githubusercontent.com/ultralytics/assets/main/social/logo-transparent.png" width="3%"/>
-      <a href="https://disuss.roboflow.com">
+      <a href="https://discuss.roboflow.com">
           <img
             src="https://media.roboflow.com/notebooks/template/icons/purple/forum.png?ik-sdk-version=javascript-1.4.3&updatedAt=1672949633584"
             width="3%"

docs/changelog.md

@@ -1,3 +1,365 @@
+### 0.23.0 <small>Aug 28, 2024</small>
+
+- Added [#930](https://github.com/roboflow/supervision/pull/930): `IconAnnotator`, a [new annotator](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.IconAnnotator) that allows drawing icons on each detection. Useful if you want to draw a specific icon for each class.
+
+```python
+import supervision as sv
+from inference import get_model
+
+image = <SOURCE_IMAGE_PATH>
+icon_dog = <DOG_PNG_PATH>
+icon_cat = <CAT_PNG_PATH>
+
+model = get_model(model_id="yolov8n-640")
+results = model.infer(image)[0]
+detections = sv.Detections.from_inference(results)
+
+icon_paths = []
+for class_name in detections.data["class_name"]:
+    if class_name == "dog":
+        icon_paths.append(icon_dog)
+    elif class_name == "cat":
+        icon_paths.append(icon_cat)
+    else:
+        icon_paths.append("")
+
+icon_annotator = sv.IconAnnotator()
+annotated_frame = icon_annotator.annotate(
+    scene=image.copy(),
+    detections=detections,
+    icon_path=icon_paths
+)
+```
+
+- Added [#1385](https://github.com/roboflow/supervision/pull/1385): [`BackgroundColorAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.BackgroundColorAnnotator), that draws an overlay on the background images of the detections.
+
+```python
+import supervision as sv
+from inference import get_model
+
+image = <SOURCE_IMAGE_PATH>
+
+model = get_model(model_id="yolov8n-640")
+results = model.infer(image)[0]
+detections = sv.Detections.from_inference(results)
+
+background_overlay_annotator = sv.BackgroundOverlayAnnotator()
+annotated_frame = background_overlay_annotator.annotate(
+    scene=image.copy(),
+    detections=detections
+)
+```
+
+- Added [#1386](https://github.com/roboflow/supervision/pull/1386): Support for Transformers v5 functions in [`sv.Detections.from_transformers`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_transformers). This includes the `DetrImageProcessor` methods `post_process_object_detection`, `post_process_panoptic_segmentation`, `post_process_semantic_segmentation`, and `post_process_instance_segmentation`.
+
+```python
+import torch
+import supervision as sv
+from PIL import Image
+from transformers import DetrImageProcessor, DetrForObjectDetection
+
+processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50")
+model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50")
+
+image = Image.open(<SOURCE_IMAGE_PATH>)
+inputs = processor(images=image, return_tensors="pt")
+
+with torch.no_grad():
+    outputs = model(**inputs)
+
+width, height = image.size
+target_size = torch.tensor([[height, width]])
+results = processor.post_process_object_detection(
+    outputs=outputs, target_sizes=target_size)[0]
+detections = sv.Detections.from_transformers(
+    transformers_results=results,
+    id2label=model.config.id2label)
+```
+
+- Added [#1354](https://github.com/roboflow/supervision/pull/1354): Ultralytics SAM (Segment Anything Model) support in [`sv.Detections.from_ultralytics`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_ultralytics). [SAM2](https://sam2.metademolab.com/) was released during this update, and is already supported via [`sv.Detections.from_sam`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_sam).
+
+```python
+import supervision as sv
+from segment_anything import (
+    sam_model_registry,
+    SamAutomaticMaskGenerator
+)
+
+sam_model_reg = sam_model_registry[MODEL_TYPE]
+sam = sam_model_reg(checkpoint=CHECKPOINT_PATH).to(device=DEVICE)
+mask_generator = SamAutomaticMaskGenerator(sam)
+sam_result = mask_generator.generate(IMAGE)
+detections = sv.Detections.from_sam(sam_result=sam_result)
+```
+
+- Added [#1458](https://github.com/roboflow/supervision/pull/1458): `outline_color` options for [`TriangleAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.TriangleAnnotator) and [`DotAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.DotAnnotator).
+
+- Added [#1409](https://github.com/roboflow/supervision/pull/1409): `text_color` option for [`VertexLabelAnnotator`](https://supervision.roboflow.com/latest/keypoint/annotators/#supervision.keypoint.annotators.VertexLabelAnnotator) keypoint annotator.
+
+- Changed [#1434](https://github.com/roboflow/supervision/pull/1434): [`InferenceSlicer`](https://supervision.roboflow.com/latest/detection/tools/inference_slicer/) now features an `overlap_ratio_wh` parameter, making it easier to compute slice sizes when handling overlapping slices.
+
+- Fix [#1448](https://github.com/roboflow/supervision/pull/1448): Various annotator type issues have been resolved, supporting expanded error handling.
+
+- Fix [#1348](https://github.com/roboflow/supervision/pull/1348): Introduced a new method for [seeking to a specific video frame](https://supervision.roboflow.com/latest/utils/video/#supervision.utils.video.get_video_frames_generator), addressing cases where traditional seek methods were failing. It can be enabled with `iterative_seek=True`.
+
+```python
+import supervision as sv
+
+for frame in sv.get_video_frames_generator(
+    source_path=<SOURCE_VIDEO_PATH>,
+    start=60,
+    iterative_seek=True
+):
+    ...
+```
+
+- Fix [#1424](https://github.com/roboflow/supervision/pull/1424): `plot_image` function now clearly indicates that the size is in inches.
+
+!!! failure "Removed"
+
+    The `track_buffer`, `track_thresh`, and `match_thresh` parameters in [`ByteTrack`](trackers.md/#supervision.tracker.byte_tracker.core.ByteTrack) are deprecated and were removed as of `supervision-0.23.0`. Use `lost_track_buffer,` `track_activation_threshold`, and `minimum_matching_threshold` instead.
+
+!!! failure "Removed"
+
+    The `triggering_position ` parameter in [`sv.PolygonZone`](detection/tools/polygon_zone.md/#supervision.detection.tools.polygon_zone.PolygonZone) was removed as of `supervision-0.23.0`. Use `triggering_anchors ` instead.
+
+!!! failure "Deprecated"
+
+    `overlap_filter_strategy` in `InferenceSlicer.__init__` is deprecated and will be removed in `supervision-0.27.0`. Use `overlap_strategy` instead.
+
+!!! failure "Deprecated"
+
+    `overlap_ratio_wh` in `InferenceSlicer.__init__` is deprecated and will be removed in `supervision-0.27.0`. Use `overlap_wh` instead.
+
+### 0.22.0 <small>Jul 12, 2024</small>
+
+- Added [#1326](https://github.com/roboflow/supervision/pull/1326): [`sv.DetectionsDataset`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset) and [`sv.ClassificationDataset`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.ClassificationDataset) allowing to load the images into memory only when necessary (lazy loading).
+
+!!! failure "Deprecated"
+
+    Constructing `DetectionDataset` with parameter `images` as `Dict[str, np.ndarray]` is deprecated and will be removed in `supervision-0.26.0`. Please pass a list of paths `List[str]` instead.
+
+!!! failure "Deprecated"
+
+    The `DetectionDataset.images` property is deprecated and will be removed in `supervision-0.26.0`. Please loop over images with `for path, image, annotation in dataset:`, as that does not require loading all images into memory.
+
+```python
+import roboflow
+from roboflow import Roboflow
+import supervision as sv
+
+roboflow.login()
+rf = Roboflow()
+
+project = rf.workspace(<WORKSPACE_ID>).project(<PROJECT_ID>)
+dataset = project.version(<PROJECT_VERSION>).download("coco")
+
+ds_train = sv.DetectionDataset.from_coco(
+    images_directory_path=f"{dataset.location}/train",
+    annotations_path=f"{dataset.location}/train/_annotations.coco.json",
+)
+
+path, image, annotation = ds_train[0]
+    # loads image on demand
+
+for path, image, annotation in ds_train:
+    # loads image on demand
+```
+
+- Added [#1296](https://github.com/roboflow/supervision/pull/1296): [`sv.Detections.from_lmm`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_lmm) now supports parsing results from the [Florence 2](https://huggingface.co/microsoft/Florence-2-large) model, extending the capability to handle outputs from this Large Multimodal Model (LMM). This includes detailed object detection, OCR with region proposals, segmentation, and more. Find out more in our [Colab notebook](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/how-to-finetune-florence-2-on-detection-dataset.ipynb).
+
+- Added [#1232](https://github.com/roboflow/supervision/pull/1232) to support keypoint detection with Mediapipe. Both [legacy](https://colab.research.google.com/github/googlesamples/mediapipe/blob/main/examples/pose_landmarker/python/%5BMediaPipe_Python_Tasks%5D_Pose_Landmarker.ipynb) and [modern](https://ai.google.dev/edge/mediapipe/solutions/vision/pose_landmarker/python) pipelines are supported. See [`sv.KeyPoints.from_mediapipe`](https://supervision.roboflow.com/latest/keypoint/core/#supervision.keypoint.core.KeyPoints.from_mediapipe) for more.
+
+- Added [#1316](https://github.com/roboflow/supervision/pull/1316): [`sv.KeyPoints.from_mediapipe`](https://supervision.roboflow.com/latest/keypoint/core/#supervision.keypoint.core.KeyPoints.from_mediapipe) extended to support FaceMesh from Mediapipe. This enhancement allows for processing both face landmarks from `FaceLandmarker`, and legacy results from `FaceMesh`.
+
+- Added [#1310](https://github.com/roboflow/supervision/pull/1310): [`sv.KeyPoints.from_detectron2`](https://supervision.roboflow.com/latest/keypoint/core/#supervision.keypoint.core.KeyPoints.from_detectron2) is a new `KeyPoints` method, adding support for extracting keypoints from the popular [Detectron 2](https://github.com/facebookresearch/detectron2) platform.
+
+- Added [#1300](https://github.com/roboflow/supervision/pull/1300): [`sv.Detections.from_detectron2`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_detectron2) now supports segmentation models detectron2. The resulting masks can be used with [`sv.MaskAnnotator`](https://supervision.roboflow.com/latest/annotators/#supervision.annotators.core.MaskAnnotator) for displaying annotations.
+
+```python
+import supervision as sv
+from detectron2 import model_zoo
+from detectron2.engine import DefaultPredictor
+from detectron2.config import get_cfg
+import cv2
+
+image = cv2.imread(<SOURCE_IMAGE_PATH>)
+cfg = get_cfg()
+cfg.merge_from_file(model_zoo.get_config_file("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml"))
+cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")
+predictor = DefaultPredictor(cfg)
+
+result = predictor(image)
+detections = sv.Detections.from_detectron2(result)
+
+mask_annotator = sv.MaskAnnotator()
+annotated_frame = mask_annotator.annotate(scene=image.copy(), detections=detections)
+```
+
+- Added [#1277](https://github.com/roboflow/supervision/pull/1277): if you provide a font that supports symbols of a language, [`sv.RichLabelAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.LabelAnnotator.annotate) will draw them on your images.
+  - Various other annotators have been revised to ensure proper in-place functionality when used with `numpy` arrays. Additionally, we fixed a bug where `sv.ColorAnnotator` was filling boxes with solid color when used in-place.
+
+```python
+import cv2
+import supervision as sv
+import
+
+image = cv2.imread(<SOURCE_IMAGE_PATH>)
+
+model = get_model(model_id="yolov8n-640")
+results = model.infer(image)[0]
+detections = sv.Detections.from_inference(results)
+
+rich_label_annotator = sv.RichLabelAnnotator(font_path=<TTF_FONT_PATH>)
+annotated_image = rich_label_annotator.annotate(scene=image.copy(), detections=detections)
+```
+
+- Added [#1227](https://github.com/roboflow/supervision/pull/1227): Added support for loading Oriented Bounding Boxes dataset in YOLO format.
+
+```python
+import supervision as sv
+
+train_ds = sv.DetectionDataset.from_yolo(
+    images_directory_path="/content/dataset/train/images",
+    annotations_directory_path="/content/dataset/train/labels",
+    data_yaml_path="/content/dataset/data.yaml",
+    is_obb=True
+)
+
+_, image, detections in train_ds[0]
+
+obb_annotator = OrientedBoxAnnotator()
+annotated_image = obb_annotator.annotate(scene=image.copy(), detections=detections)
+```
+
+- Fixed [#1312](https://github.com/roboflow/supervision/pull/1312): Fixed [`CropAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.TraceAnnotator.annotate).
+
+!!! failure "Removed"
+
+    `BoxAnnotator` was removed, however `BoundingBoxAnnotator` has been renamed to `BoxAnnotator`. Use a combination of [`BoxAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.BoxAnnotator) and [`LabelAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.LabelAnnotator) to simulate old `BoundingBox` behavior.
+
+!!! failure "Deprecated"
+
+    The name `BoundingBoxAnnotator` has been deprecated and will be removed in `supervision-0.26.0`. It has been renamed to [`BoxAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.BoxAnnotator).
+
+- Added [#975](https://github.com/roboflow/supervision/pull/975) 📝 New Cookbooks: serialize detections into [json](https://github.com/roboflow/supervision/blob/de896189b83a1f9434c0a37dd9192ee00d2a1283/docs/notebooks/serialise-detections-to-json.ipynb) and [csv](https://github.com/roboflow/supervision/blob/de896189b83a1f9434c0a37dd9192ee00d2a1283/docs/notebooks/serialise-detections-to-csv.ipynb).
+
+- Added [#1290](https://github.com/roboflow/supervision/pull/1290): Mostly an internal change, our file utility function now support both `str` and `pathlib` paths.
+
+- Added [#1340](https://github.com/roboflow/supervision/pull/1340): Two new methods for converting between bounding box formats - [`xywh_to_xyxy`](https://supervision.roboflow.com/latest/detection/utils/#supervision.detection.utils.xywh_to_xyxy) and [`xcycwh_to_xyxy`](https://supervision.roboflow.com/latest/detection/utils/#supervision.detection.utils.xcycwh_to_xyxy)
+
+!!! failure "Removed"
+
+    `from_roboflow` method has been removed due to deprecation. Use [from_inference](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_inference) instead.
+
+!!! failure "Removed"
+
+    `Color.white()` has been removed due to deprecation. Use `color.WHITE` instead.
+
+!!! failure "Removed"
+
+    `Color.black()` has been removed due to deprecation. Use `color.BLACK` instead.
+
+!!! failure "Removed"
+
+    `Color.red()` has been removed due to deprecation. Use `color.RED` instead.
+
+!!! failure "Removed"
+
+    `Color.green()` has been removed due to deprecation. Use `color.GREEN` instead.
+
+!!! failure "Removed"
+
+    `Color.blue()` has been removed due to deprecation. Use `color.BLUE` instead.
+
+!!! failure "Removed"
+
+    `ColorPalette.default()` has been removed due to deprecation. Use [ColorPalette.DEFAULT](https://supervision.roboflow.com/latest/utils/draw/#supervision.draw.color.ColorPalette.DEFAULT) instead.
+
+!!! failure "Removed"
+
+    `FPSMonitor.__call__` has been removed due to deprecation. Use the attribute [FPSMonitor.fps](https://supervision.roboflow.com/latest/utils/video/#supervision.utils.video.FPSMonitor.fps) instead.
+
+### 0.21.0 <small>Jun 5, 2024</small>
+
+- Added [#500](https://github.com/roboflow/supervision/pull/500): [`sv.Detections.with_nmm`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.with_nmm) to perform non-maximum merging on the current set of object detections.
+
+- Added [#1221](https://github.com/roboflow/supervision/pull/1221): [`sv.Detections.from_lmm`](https://supervision.roboflow.com/latest/detection/core/#supervision.detection.core.Detections.from_lmm) allowing to parse Large Multimodal Model (LMM) text result into [`sv.Detections`](https://supervision.roboflow.com/latest/detection/core/) object. For now `from_lmm` supports only [PaliGemma](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/how-to-finetune-paligemma-on-detection-dataset.ipynb) result parsing.
+
+```python
+import supervision as sv
+
+paligemma_result = "<loc0256><loc0256><loc0768><loc0768> cat"
+detections = sv.Detections.from_lmm(
+    sv.LMM.PALIGEMMA,
+    paligemma_result,
+    resolution_wh=(1000, 1000),
+    classes=['cat', 'dog']
+)
+detections.xyxy
+# array([[250., 250., 750., 750.]])
+
+detections.class_id
+# array([0])
+```
+
+- Added [#1236](https://github.com/roboflow/supervision/pull/1236): [`sv.VertexLabelAnnotator`](https://supervision.roboflow.com/latest/keypoint/annotators/#supervision.keypoint.annotators.EdgeAnnotator.annotate) allowing to annotate every vertex of a keypoint skeleton with custom text and color.
+
+```python
+import supervision as sv
+
+image = ...
+key_points = sv.KeyPoints(...)
+
+edge_annotator = sv.EdgeAnnotator(
+    color=sv.Color.GREEN,
+    thickness=5
+)
+annotated_frame = edge_annotator.annotate(
+    scene=image.copy(),
+    key_points=key_points
+)
+```
+
+- Added [#1147](https://github.com/roboflow/supervision/pull/1147): [`sv.KeyPoints.from_inference`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints.from_inference) allowing to create [`sv.KeyPoints`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints) from [Inference](https://github.com/roboflow/inference) result.
+
+- Added [#1138](https://github.com/roboflow/supervision/pull/1138): [`sv.KeyPoints.from_yolo_nas`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints.from_yolo_nas) allowing to create [`sv.KeyPoints`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints) from [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) result.
+
+- Added [#1163](https://github.com/roboflow/supervision/pull/1163): [`sv.mask_to_rle`](https://supervision.roboflow.com/develop/datasets/utils/#supervision.dataset.utils.rle_to_mask) and [`sv.rle_to_mask`](https://supervision.roboflow.com/develop/datasets/utils/#supervision.dataset.utils.rle_to_mask) allowing for easy conversion between mask and rle formats.
+
+- Changed [#1236](https://github.com/roboflow/supervision/pull/1236): [`sv.InferenceSlicer`](https://supervision.roboflow.com/develop/detection/tools/inference_slicer/) allowing to select overlap filtering strategy (`NONE`, `NON_MAX_SUPPRESSION` and `NON_MAX_MERGE`).
+
+- Changed [#1178](https://github.com/roboflow/supervision/pull/1178): [`sv.InferenceSlicer`](https://supervision.roboflow.com/develop/detection/tools/inference_slicer/) adding instance segmentation model support.
+
+```python
+import cv2
+import numpy as np
+import supervision as sv
+from inference import get_model
+
+model = get_model(model_id="yolov8x-seg-640")
+image = cv2.imread(<SOURCE_IMAGE_PATH>)
+
+def callback(image_slice: np.ndarray) -> sv.Detections:
+    results = model.infer(image_slice)[0]
+    return sv.Detections.from_inference(results)
+
+slicer = sv.InferenceSlicer(callback = callback)
+detections = slicer(image)
+
+mask_annotator = sv.MaskAnnotator()
+label_annotator = sv.LabelAnnotator()
+
+annotated_image = mask_annotator.annotate(
+    scene=image, detections=detections)
+annotated_image = label_annotator.annotate(
+    scene=annotated_image, detections=detections)
+```
+
+- Changed [#1228](https://github.com/roboflow/supervision/pull/1228): [`sv.LineZone`](https://supervision.roboflow.com/develop/detection/tools/line_zone/) making it 10-20 times faster, depending on the use case.
+
+- Changed [#1163](https://github.com/roboflow/supervision/pull/1163): [`sv.DetectionDataset.from_coco`](https://supervision.roboflow.com/develop/datasets/core/#supervision.dataset.core.DetectionDataset.from_coco) and [`sv.DetectionDataset.as_coco`](https://supervision.roboflow.com/develop/datasets/core/#supervision.dataset.core.DetectionDataset.as_coco) adding support for run-length encoding (RLE) mask format.
+
 ### 0.20.0 <small>April 24, 2024</small>
 
 - Added [#1128](https://github.com/roboflow/supervision/pull/1128): [`sv.KeyPoints`](/0.20.0/keypoint/core/#supervision.keypoint.core.KeyPoints) to provide initial support for pose estimation and broader keypoint detection models.
@@ -84,8 +446,6 @@ with csv_sink:
 
 - Added [#819](https://github.com/roboflow/supervision/pull/819): [`sv.JSONSink`](/0.19.0/detection/tools/save_detections/#supervision.detection.tools.csv_sink.JSONSink) allowing for the straightforward saving of image, video, or stream inference results in a `.json` file.
 
-```python
-
 ```python
 import supervision as sv
 from ultralytics import YOLO
@@ -133,7 +493,7 @@ annotated_frame = crop_annotator.annotate(
 
 !!! failure "Deprecated"
 
-    The `track_buffer`, `track_thresh`, and `match_thresh` parameters in `sv.ByterTrack` are deprecated and will be removed in `supervision-0.23.0`. Use `lost_track_buffer,` `track_activation_threshold`, and `minimum_matching_threshold` instead.
+    The `track_buffer`, `track_thresh`, and `match_thresh` parameters in `sv.ByteTrack` are deprecated and will be removed in `supervision-0.23.0`. Use `lost_track_buffer,` `track_activation_threshold`, and `minimum_matching_threshold` instead.
 
 - Changed [#910](https://github.com/roboflow/supervision/pull/910): [`sv.PolygonZone`](/0.19.0/detection/tools/polygon_zone/#supervision.detection.tools.polygon_zone.PolygonZone) to now accept a list of specific box anchors that must be in zone for a detection to be counted.
 
@@ -205,7 +565,6 @@ ColorPalette(colors=[Color(r=68, g=1, b=84), Color(r=59, g=82, b=139), ...])
 
     `sv.ColorPalette.default()` is deprecated and will be removed in `supervision-0.22.0`. Use `sv.ColorPalette.DEFAULT` instead.
 
-
 - Changed [#769](https://github.com/roboflow/supervision/pull/769): [`sv.ColorPalette.DEFAULT`](/0.18.0/draw/color/#colorpalette) value, giving users a more extensive set of annotation colors.
 
 - Changed [#677](https://github.com/roboflow/supervision/pull/677): `sv.Detections.from_roboflow` to [`sv.Detections.from_inference`](/0.18.0/detection/core/#supervision.detection.core.Detections.from_inference) streamlining its functionality to be compatible with both the both [inference](https://github.com/roboflow/inference) pip package and the Robloflow [hosted API](https://docs.roboflow.com/deploy/hosted-api).
@@ -214,7 +573,6 @@ ColorPalette(colors=[Color(r=68, g=1, b=84), Color(r=59, g=82, b=139), ...])
 
     `Detections.from_roboflow()` is deprecated and will be removed in `supervision-0.22.0`. Use `Detections.from_inference` instead.
 
-
 - Fixed [#735](https://github.com/roboflow/supervision/pull/735): [`sv.LineZone`](/0.18.0/detection/tools/line_zone/#linezone) functionality to accurately update the counter when an object crosses a line from any direction, including from the side. This enhancement enables more precise tracking and analytics, such as calculating individual in/out counts for each lane on the road.
 
 ### 0.17.0 <small>December 06, 2023</small>
@@ -307,13 +665,12 @@ ColorPalette(colors=[Color(r=68, g=1, b=84), Color(r=59, g=82, b=139), ...])
 
 - Fixed [#477](https://github.com/roboflow/supervision/pull/477): Poetry env definition allowing proper local installation.
 
-- Fixed [#430](https://github.com/roboflow/supervision/pull/430):  [`sv.ByteTrack`](/0.16.0/trackers/#supervision.tracker.byte_tracker.core.ByteTrack) to return `np.array([], dtype=int)` when `svDetections` is empty.
+- Fixed [#430](https://github.com/roboflow/supervision/pull/430): [`sv.ByteTrack`](/0.16.0/trackers/#supervision.tracker.byte_tracker.core.ByteTrack) to return `np.array([], dtype=int)` when `svDetections` is empty.
 
 !!! failure "Deprecated"
 
     `sv.Detections.from_yolov8` and `sv.Classifications.from_yolov8` as those are now replaced by [`sv.Detections.from_ultralytics`](/0.16.0/detection/core/#supervision.detection.core.Detections.from_ultralytics) and [`sv.Classifications.from_ultralytics`](/0.16.0/classification/core/#supervision.classification.core.Classifications.from_ultralytics).
 
-
 ### 0.15.0 <small>October 5, 2023</small>
 
 - Added [#170](https://github.com/roboflow/supervision/pull/170): [`sv.BoundingBoxAnnotator`](/0.15.0/annotators/#supervision.annotators.core.BoundingBoxAnnotator) allowing to annotate images and videos with bounding boxes.

docs/cookbooks.md

@@ -1,7 +1,6 @@
 ---
 template: cookbooks.html
 comments: true
-status: new
 hide:
   - navigation
   - toc

docs/datasets/utils.md

@@ -0,0 +1,17 @@
+---
+comments: true
+---
+
+# Datasets Utils
+
+<div class="md-typeset">
+  <h2><a href="#supervision.dataset.utils.rle_to_mask">rle_to_mask</a></h2>
+</div>
+
+:::supervision.dataset.utils.rle_to_mask
+
+<div class="md-typeset">
+  <h2><a href="#supervision.dataset.utils.mask_to_rle">mask_to_rle</a></h2>
+</div>
+
+:::supervision.dataset.utils.mask_to_rle

docs/deprecated.md

@@ -3,17 +3,34 @@ comments: true
 status: deprecated
 ---
 
-These features are phased out due to better alternatives or potential issues in future versions. Deprecated functionalities are supported for **three subsequent releases**, providing time for users to transition to updated methods.
+# Deprecated
 
-- [`Detections.from_froboflow`](detection/core.md/#supervision.detection.core.Detections.from_roboflow) is deprecated and will be removed in `supervision-0.22.0`. Use [`Detections.from_inference`](detection/core.md/#supervision.detection.core.Detections.from_inference) instead.
-- The method `Color.white()` is deprecated and will be removed in `supervision-0.22.0`. Use the constant `Color.WHITE` instead.
-- The method `Color.black()` is deprecated and will be removed in `supervision-0.22.0`. Use the constant `Color.BLACK` instead.
-- The method `Color.red()` is deprecated and will be removed in `supervision-0.22.0`. Use the constant `Color.RED` instead.
-- The method `Color.green()` is deprecated and will be removed in `supervision-0.22.0`. Use the constant `Color.GREEN` instead.
-- The method `Color.blue()` is deprecated and will be removed in `supervision-0.22.0`. Use the constant `Color.BLUE` instead.
-- The method [`ColorPalette.default()`](draw/color.md/#supervision.draw.color.ColorPalette.default) is deprecated and will be removed in `supervision-0.22.0`. Use the constant [`ColorPalette.DEFAULT`](draw/color.md/#supervision.draw.color.ColorPalette.DEFAULT) instead.
-- `BoxAnnotator` is deprecated and will be removed in `supervision-0.22.0`. Use [`BoundingBoxAnnotator`](detection/annotators.md/#supervision.annotators.core.BoundingBoxAnnotator) and [`LabelAnnotator`](detection/annotators.md/#supervision.annotators.core.LabelAnnotator) instead.
-- The method [`FPSMonitor.__call__`](utils/video.md/#supervision.utils.video.FPSMonitor.__call__) is deprecated and will be removed in `supervision-0.22.0`. Use the attribute [`FPSMonitor.fps`](utils/video.md/#supervision.utils.video.FPSMonitor.fps) instead.
-- The `track_buffer`, `track_thresh`, and `match_thresh` parameters in [`ByterTrack`](trackers.md/#supervision.tracker.byte_tracker.core.ByteTrack) are deprecated and will be removed in `supervision-0.23.0`. Use `lost_track_buffer,` `track_activation_threshold`, and `minimum_matching_threshold` instead.
-- The `triggering_position ` parameter in [`sv.PolygonZone`](detection/tools/polygon_zone.md/#supervision.detection.tools.polygon_zone.PolygonZone) is deprecated and will be removed in `supervision-0.23.0`. Use `triggering_anchors ` instead.
-- The `frame_resolution_wh ` parameter in [`sv.PolygonZone`](detection/tools/polygon_zone.md/#supervision.detection.tools.polygon_zone.PolygonZone) is deprecated and will be removed in `supervision-0.24.0`.
+These features are phased out due to better alternatives or potential issues in future versions. Deprecated functionalities are supported for **five subsequent releases**, providing time for users to transition to updated methods.
+
+- The `frame_resolution_wh ` parameter in [`sv.PolygonZone`](detection/tools/polygon_zone.md/#supervision.detection.tools.polygon_zone.PolygonZone) will be removed in `supervision-0.24.0`.
+
+- Constructing [`DetectionDataset`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset) and [`ClassificationDataset`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.ClassificationDataset) with parameter `images` as `Dict[str, np.ndarray]` will be removed in `supervision-0.26.0`. Please pass a list of paths `List[str]` instead.
+- The `DetectionDataset.images` property will be removed in `supervision-0.26.0`. Please loop over images with `for path, image, annotation in dataset:`, as that does not require loading all images into memory.
+- `BoundingBoxAnnotator` has been renamed to `BoxAnnotator` after the old implementation of [`BoxAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.BoxAnnotator) has been removed. `BoundingBoxAnnotator` will be removed in `supervision-0.26.0`.
+
+- `overlap_filter_strategy` in [`InferenceSlicer.__init__`](https://supervision.roboflow.com/latest/detection/tools/inference_slicer/) is deprecated and will be removed in `supervision-0.27.0`. Use `overlap_strategy` instead.
+- `overlap_ratio_wh` in [`InferenceSlicer.__init__`](https://supervision.roboflow.com/latest/detection/tools/inference_slicer/) is deprecated and will be removed in `supervision-0.27.0`. Use `overlap_wh` instead.
+
+# Removed
+
+### 0.23.0
+
+- The `track_buffer`, `track_thresh`, and `match_thresh` parameters in [`ByteTrack`](trackers.md/#supervision.tracker.byte_tracker.core.ByteTrack) are deprecated and were removed as of `supervision-0.23.0`. Use `lost_track_buffer,` `track_activation_threshold`, and `minimum_matching_threshold` instead.
+- The `triggering_position ` parameter in [`sv.PolygonZone`](detection/tools/polygon_zone.md/#supervision.detection.tools.polygon_zone.PolygonZone) was removed as of `supervision-0.23.0`. Use `triggering_anchors ` instead.
+
+### 0.22.0
+
+- [`Detections.from_froboflow`](detection/core.md/#supervision.detection.core.Detections.from_roboflow) is removed as of `supervision-0.22.0`. Use [`Detections.from_inference`](detection/core.md/#supervision.detection.core.Detections.from_inference) instead.
+- The method `Color.white()` was removed as of `supervision-0.22.0`. Use the constant `Color.WHITE` instead.
+- The method `Color.black()` was removed as of `supervision-0.22.0`. Use the constant `Color.BLACK` instead.
+- The method `Color.red()` was removed as of `supervision-0.22.0`. Use the constant `Color.RED` instead.
+- The method `Color.green()` was removed as of `supervision-0.22.0`. Use the constant `Color.GREEN` instead.
+- The method `Color.blue()` was removed as of `supervision-0.22.0`. Use the constant `Color.BLUE` instead.
+- The method [`ColorPalette.default()`](draw/color.md/#supervision.draw.color.ColorPalette.default) was removed as of `supervision-0.22.0`. Use the constant [`ColorPalette.DEFAULT`](draw/color.md/#supervision.draw.color.ColorPalette.DEFAULT) instead.
+- `BoxAnnotator` was removed as of `supervision-0.22.0`, however `BoundingBoxAnnotator` was immediately renamed to `BoxAnnotator`. Use [`BoxAnnotator`](detection/annotators.md/#supervision.annotators.core.BoxAnnotator) and [`LabelAnnotator`](detection/annotators.md/#supervision.annotators.core.LabelAnnotator) instead of the old `BoxAnnotator`.
+- The method [`FPSMonitor.__call__`](utils/video.md/#supervision.utils.video.FPSMonitor.__call__) was removed as of `supervision-0.22.0`. Use the attribute [`FPSMonitor.fps`](utils/video.md/#supervision.utils.video.FPSMonitor.fps) instead.

docs/detection/annotators.md

@@ -5,7 +5,7 @@ status: new
 
 # Annotators
 
-=== "BoundingBox"
+=== "Box"
 
     ```python
     import supervision as sv
@@ -13,8 +13,8 @@ status: new
     image = ...
     detections = sv.Detections(...)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
-    annotated_frame = bounding_box_annotator.annotate(
+    box_annotator = sv.BoxAnnotator()
+    annotated_frame = box_annotator.annotate(
         scene=image.copy(),
         detections=detections
     )
@@ -285,6 +285,64 @@ status: new
 
     </div>
 
+=== "RichLabel"
+
+    ```python
+    import supervision as sv
+
+    image = ...
+    detections = sv.Detections(...)
+
+    labels = [
+        f"{class_name} {confidence:.2f}"
+        for class_name, confidence
+        in zip(detections['class_name'], detections.confidence)
+    ]
+
+    rich_label_annotator = sv.RichLabelAnnotator(
+        font_path="<TTF_FONT_PATH>",
+        text_position=sv.Position.CENTER
+    )
+    annotated_frame = rich_label_annotator.annotate(
+        scene=image.copy(),
+        detections=detections,
+        labels=labels
+    )
+    ```
+
+    <div class="result" markdown>
+
+    ![label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/label-annotator-example-purple.png){ align=center width="800" }
+
+    </div>
+
+=== "Icon"
+
+    ```python
+    import supervision as sv
+
+    image = ...
+    detections = sv.Detections(...)
+
+    icon_paths = [
+        "<ICON_PATH>"
+        for _ in detections
+    ]
+
+    icon_annotator = sv.IconAnnotator()
+    annotated_frame = icon_annotator.annotate(
+        scene=image.copy(),
+        detections=detections,
+        icon_path=icon_paths
+    )
+    ```
+
+    <div class="result" markdown>
+
+    ![icon-annotator-example](https://media.roboflow.com/supervision-annotator-examples/icon-annotator-example.png){ align=center width="800" }
+
+    </div>
+
 === "Crop"
 
     ```python
@@ -353,7 +411,7 @@ status: new
     trace_annotator = sv.TraceAnnotator()
 
     video_info = sv.VideoInfo.from_video_path(video_path='...')
-    frames_generator = get_video_frames_generator(source_path='...')
+    frames_generator = sv.get_video_frames_generator(source_path='...')
     tracker = sv.ByteTrack()
 
     with sv.VideoSink(target_path='...', video_info=video_info) as sink:
@@ -384,7 +442,7 @@ status: new
     heat_map_annotator = sv.HeatMapAnnotator()
 
     video_info = sv.VideoInfo.from_video_path(video_path='...')
-    frames_generator = get_video_frames_generator(source_path='...')
+    frames_generator = sv.get_video_frames_generator(source_path='...')
 
     with sv.VideoSink(target_path='...', video_info=video_info) as sink:
         for frame in frames_generator:
@@ -402,11 +460,32 @@ status: new
 
     </div>
 
+=== "Background Color"
+
+    ```python
+    import supervision as sv
+
+    image = ...
+    detections = sv.Detections(...)
+
+    background_overlay_annotator = sv.BackgroundOverlayAnnotator()
+    annotated_frame = background_overlay_annotator.annotate(
+        scene=image.copy(),
+        detections=detections
+    )
+    ```
+
+    <div class="result" markdown>
+
+    ![background-overlay-annotator-example](https://media.roboflow.com/supervision-annotator-examples/background-color-annotator-example-purple.png)
+
+    </div>
+
 <div class="md-typeset">
-    <h2><a href="#supervision.annotators.core.BoundingBoxAnnotator">BoundingBoxAnnotator</a></h2>
+    <h2><a href="#supervision.annotators.core.BoxAnnotator">BoxAnnotator</a></h2>
 </div>
 
-:::supervision.annotators.core.BoundingBoxAnnotator
+:::supervision.annotators.core.BoxAnnotator
 
 <div class="md-typeset">
     <h2><a href="#supervision.annotators.core.RoundBoxAnnotator">RoundBoxAnnotator</a></h2>
@@ -492,6 +571,18 @@ status: new
 
 :::supervision.annotators.core.LabelAnnotator
 
+<div class="md-typeset">
+    <h2><a href="#supervision.annotators.core.RichLabelAnnotator">RichLabelAnnotator</a></h2>
+</div>
+
+:::supervision.annotators.core.RichLabelAnnotator
+
+<div class="md-typeset">
+    <h2><a href="#supervision.annotators.core.IconAnnotator">IconAnnotator</a></h2>
+</div>
+
+:::supervision.annotators.core.IconAnnotator
+
 <div class="md-typeset">
     <h2><a href="#supervision.annotators.core.BlurAnnotator">BlurAnnotator</a></h2>
 </div>
@@ -516,6 +607,12 @@ status: new
 
 :::supervision.annotators.core.CropAnnotator
 
+<div class="md-typeset">
+    <h2><a href="#supervision.annotators.core.BackgroundOverlayAnnotator">BackgroundOverlayAnnotator</a></h2>
+</div>
+
+:::supervision.annotators.core.BackgroundOverlayAnnotator
+
 <div class="md-typeset">
     <h2><a href="#supervision.annotators.core.ColorLookup">ColorLookup</a></h2>
 </div>

docs/detection/double_detection_filter.md

@@ -0,0 +1,29 @@
+---
+comments: true
+---
+
+# Double Detection Filter
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.OverlapFilter">OverlapFilter</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.OverlapFilter
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.box_non_max_suppression">box_non_max_suppression</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.box_non_max_suppression
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.mask_non_max_suppression">mask_non_max_suppression</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.mask_non_max_suppression
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.box_non_max_merge">box_non_max_merge</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.box_non_max_merge

docs/detection/metrics.md

@@ -2,7 +2,10 @@
 comments: true
 ---
 
-# Metrics
+# Legacy Metrics
+
+Starting with `0.23.0`, a new metrics module is being introduced to supervision.
+Metrics here are part of the legacy evaluation API and will be deprecated in the future.
 
 <div class="md-typeset">
     <h2><a href="#supervision.metrics.detection.ConfusionMatrix">ConfusionMatrix</a></h2>

docs/detection/tools/inference_slicer.md

@@ -1,5 +1,6 @@
 ---
 comments: true
+status: new
 ---
 
 # InferenceSlicer

docs/detection/tools/save_detections.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Save Detections

docs/detection/utils.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Detection Utils
@@ -17,18 +16,6 @@ status: new
 
 :::supervision.detection.utils.mask_iou_batch
 
-<div class="md-typeset">
-  <h2><a href="#supervision.detection.utils.box_non_max_suppression">box_non_max_suppression</a></h2>
-</div>
-
-:::supervision.detection.utils.box_non_max_suppression
-
-<div class="md-typeset">
-  <h2><a href="#supervision.detection.utils.mask_non_max_suppression">mask_non_max_suppression</a></h2>
-</div>
-
-:::supervision.detection.utils.mask_non_max_suppression
-
 <div class="md-typeset">
   <h2><a href="#supervision.detection.utils.polygon_to_mask">polygon_to_mask</a></h2>
 </div>
@@ -65,8 +52,50 @@ status: new
 
 :::supervision.detection.utils.move_boxes
 
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.move_masks">move_masks</a></h2>
+</div>
+
+:::supervision.detection.utils.move_masks
+
 <div class="md-typeset">
   <h2><a href="#supervision.detection.utils.scale_boxes">scale_boxes</a></h2>
 </div>
 
 :::supervision.detection.utils.scale_boxes
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.clip_boxes">clip_boxes</a></h2>
+</div>
+
+:::supervision.detection.utils.clip_boxes
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.pad_boxes">pad_boxes</a></h2>
+</div>
+
+:::supervision.detection.utils.pad_boxes
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.xywh_to_xyxy">xywh_to_xyxy</a></h2>
+</div>
+
+:::supervision.detection.utils.xywh_to_xyxy
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.xcycwh_to_xyxy">xcycwh_to_xyxy</a></h2>
+</div>
+
+:::supervision.detection.utils.xcycwh_to_xyxy
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.contains_holes">contains_holes</a></h2>
+</div>
+
+:::supervision.detection.utils.contains_holes
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.contains_multiple_segments">contains_multiple_segments</a></h2>
+</div>
+
+:::supervision.detection.utils.contains_multiple_segments

docs/how_to/detect_and_annotate.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Detect and Annotate
@@ -136,7 +135,7 @@ You can load predictions from other computer vision frameworks and libraries usi
 
 ## Annotate Image with Detections
 
-Finally, we can annotate the image with the predictions. Since we are working with an object detection model, we will use the [`sv.BoundingBoxAnnotator`](/latest/annotators/#supervision.annotators.core.BoundingBoxAnnotator) and [`sv.LabelAnnotator`](/latest/annotators/#supervision.annotators.core.LabelAnnotator) classes.
+Finally, we can annotate the image with the predictions. Since we are working with an object detection model, we will use the [`sv.BoxAnnotator`](/latest/annotators/#supervision.annotators.core.BoxAnnotator) and [`sv.LabelAnnotator`](/latest/annotators/#supervision.annotators.core.LabelAnnotator) classes.
 
 === "Inference"
 
@@ -150,10 +149,10 @@ Finally, we can annotate the image with the predictions. Since we are working wi
     results = model.infer(image)[0]
     detections = sv.Detections.from_inference(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -171,10 +170,10 @@ Finally, we can annotate the image with the predictions. Since we are working wi
     results = model(image)[0]
     detections = sv.Detections.from_ultralytics(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -205,10 +204,10 @@ Finally, we can annotate the image with the predictions. Since we are working wi
         transformers_results=results,
         id2label=model.config.id2label)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -234,7 +233,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
     results = model.infer(image)[0]
     detections = sv.Detections.from_inference(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     labels = [
@@ -243,7 +242,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
         in zip(detections['class_name'], detections.confidence)
     ]
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections, labels=labels)
@@ -261,7 +260,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
     results = model(image)[0]
     detections = sv.Detections.from_ultralytics(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     labels = [
@@ -270,7 +269,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
         in zip(detections['class_name'], detections.confidence)
     ]
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections, labels=labels)
@@ -301,7 +300,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
         transformers_results=results,
         id2label=model.config.id2label)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     labels = [
@@ -310,7 +309,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
         in zip(detections['class_name'], detections.confidence)
     ]
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections, labels=labels)
@@ -323,7 +322,7 @@ override this behavior by passing a list of custom `labels` to the `annotate` me
 If you are running the segmentation model
 [`sv.MaskAnnotator`](/latest/annotators/#supervision.annotators.core.MaskAnnotator)
 is a drop-in replacement for
-[`sv.BoundingBoxAnnotator`](/latest/annotators/#supervision.annotators.core.BoundingBoxAnnotator)
+[`sv.BoxAnnotator`](/latest/annotators/#supervision.annotators.core.BoxAnnotator)
 that will allow you to draw masks instead of boxes.
 
 === "Inference"

docs/how_to/detect_small_objects.md

@@ -1,12 +1,11 @@
 ---
 comments: true
-status: new
 ---
 
 # Detect Small Objects
 
 This guide shows how to detect small objects
-with the  [Inference](https://github.com/roboflow/inference),
+with the [Inference](https://github.com/roboflow/inference),
 [Ultralytics](https://github.com/ultralytics/ultralytics) or
 [Transformers](https://github.com/huggingface/transformers) packages using
 [`InferenceSlicer`](/latest/detection/tools/inference_slicer/#supervision.detection.tools.inference_slicer.InferenceSlicer).
@@ -32,10 +31,10 @@ size relative to the image resolution.
     results = model.infer(image)[0]
     detections = sv.Detections.from_inference(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -53,10 +52,10 @@ size relative to the image resolution.
     results = model(image)[0]
     detections = sv.Detections.from_ultralytics(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -68,10 +67,10 @@ size relative to the image resolution.
     import torch
     import supervision as sv
     from PIL import Image
-    from transformers import DetrImageProcessor, DetrForObjectDetection
+    from transformers import DetrImageProcessor, DetrForSegmentation
 
     processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50")
-    model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50")
+    model = DetrForSegmentation.from_pretrained("facebook/detr-resnet-50")
 
     image = Image.open(<SOURCE_IMAGE_PATH>)
     inputs = processor(images=image, return_tensors="pt")
@@ -79,13 +78,13 @@ size relative to the image resolution.
     with torch.no_grad():
         outputs = model(**inputs)
 
-    width, height = image.size
-    target_size = torch.tensor([[height, width]])
+    width, height = image_slice.size
+    target_size = torch.tensor([[width, height]])
     results = processor.post_process_object_detection(
         outputs=outputs, target_sizes=target_size)[0]
     detections = sv.Detections.from_transformers(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     labels = [
@@ -94,7 +93,7 @@ size relative to the image resolution.
         in detections.class_id
     ]
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections, labels=labels)
@@ -120,10 +119,10 @@ is less effective for ultra-high-resolution images (4K and above).
     results = model.infer(image)[0]
     detections = sv.Detections.from_inference(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -141,10 +140,10 @@ is less effective for ultra-high-resolution images (4K and above).
     results = model(image, imgsz=1280)[0]
     detections = sv.Detections.from_ultralytics(results)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -175,15 +174,15 @@ objects within each, and aggregating the results.
 
     def callback(image_slice: np.ndarray) -> sv.Detections:
         results = model.infer(image_slice)[0]
-        detections = sv.Detections.from_inference(results)
+        return sv.Detections.from_inference(results)
 
     slicer = sv.InferenceSlicer(callback = callback)
     detections = slicer(image)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -207,10 +206,10 @@ objects within each, and aggregating the results.
     slicer = sv.InferenceSlicer(callback = callback)
     detections = slicer(image)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections)
@@ -239,8 +238,8 @@ objects within each, and aggregating the results.
         with torch.no_grad():
             outputs = model(**inputs)
 
-        width, height = image.size
-        target_size = torch.tensor([[height, width]])
+        width, height = image_slice.size
+        target_size = torch.tensor([[width, height]])
         results = processor.post_process_object_detection(
             outputs=outputs, target_sizes=target_size)[0]
         return sv.Detections.from_transformers(results)
@@ -248,7 +247,7 @@ objects within each, and aggregating the results.
     slicer = sv.InferenceSlicer(callback = callback)
     detections = slicer(image)
 
-    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     labels = [
@@ -257,10 +256,70 @@ objects within each, and aggregating the results.
         in detections.class_id
     ]
 
-    annotated_image = bounding_box_annotator.annotate(
+    annotated_image = box_annotator.annotate(
         scene=image, detections=detections)
     annotated_image = label_annotator.annotate(
         scene=annotated_image, detections=detections, labels=labels)
     ```
 
 ![detection-with-inference-slicer](https://media.roboflow.com/supervision_detect_small_objects_example_3.png)
+
+## Small Object Segmentation
+
+[`InferenceSlicer`](/latest/detection/tools/inference_slicer/#supervision.detection.tools.inference_slicer.InferenceSlicer) can perform segmentation tasks too.
+
+=== "Inference"
+
+    ```{ .py hl_lines="6 16 19-20" }
+    import cv2
+    import numpy as np
+    import supervision as sv
+    from inference import get_model
+
+    model = get_model(model_id="yolov8x-seg-640")
+    image = cv2.imread(<SOURCE_IMAGE_PATH>)
+
+    def callback(image_slice: np.ndarray) -> sv.Detections:
+        results = model.infer(image_slice)[0]
+        return sv.Detections.from_inference(results)
+
+    slicer = sv.InferenceSlicer(callback = callback)
+    detections = slicer(image)
+
+    mask_annotator = sv.MaskAnnotator()
+    label_annotator = sv.LabelAnnotator()
+
+    annotated_image = mask_annotator.annotate(
+        scene=image, detections=detections)
+    annotated_image = label_annotator.annotate(
+        scene=annotated_image, detections=detections)
+    ```
+
+=== "Ultralytics"
+
+    ```{ .py hl_lines="6 16 19-20" }
+    import cv2
+    import numpy as np
+    import supervision as sv
+    from ultralytics import YOLO
+
+    model = YOLO("yolov8x-seg.pt")
+    image = cv2.imread(<SOURCE_IMAGE_PATH>)
+
+    def callback(image_slice: np.ndarray) -> sv.Detections:
+        result = model(image_slice)[0]
+        return sv.Detections.from_ultralytics(result)
+
+    slicer = sv.InferenceSlicer(callback = callback)
+    detections = slicer(image)
+
+    mask_annotator = sv.MaskAnnotator()
+    label_annotator = sv.LabelAnnotator()
+
+    annotated_image = mask_annotator.annotate(
+        scene=image, detections=detections)
+    annotated_image = label_annotator.annotate(
+        scene=annotated_image, detections=detections)
+    ```
+
+![detection-with-inference-slicer](https://media.roboflow.com/supervision-docs/inference-slicer-segmentation-example.png)

docs/how_to/process_datasets.md

@@ -0,0 +1,454 @@
+---
+comments: true
+---
+
+With Supervision, you can load and manipulate classification, object detection, and
+segmentation datasets. This tutorial will walk you through how to load, split, merge,
+visualize, and augment datasets in Supervision.
+
+## Download Dataset
+
+In this tutorial, we will use a dataset from
+[Roboflow Universe](https://universe.roboflow.com/), a public repository of
+thousands of computer vision datasets. If you already have your dataset in
+[COCO](https://roboflow.com/formats/coco-json),
+[YOLO](https://roboflow.com/formats/yolov8-pytorch-txt),
+or [Pascal VOC](https://roboflow.com/formats/pascal-voc-xml) format, you can skip this
+section.
+
+```bash
+pip install roboflow
+```
+
+Next, log into your Roboflow account and download the dataset of your choice in the
+COCO, YOLO, or Pascal VOC format. You can customize the following code snippet with
+your workspace ID, project ID, and version number.
+
+=== "COCO"
+
+    ```python
+    import roboflow
+
+    roboflow.login()
+
+    rf = roboflow.Roboflow()
+    project = rf.workspace('<WORKSPACE_ID>').project('<PROJECT_ID>')
+    dataset = project.version('<PROJECT_VERSION>').download("coco")
+    ```
+
+=== "YOLO"
+
+    ```python
+    import roboflow
+
+    roboflow.login()
+
+    rf = roboflow.Roboflow()
+    project = rf.workspace('<WORKSPACE_ID>').project('<PROJECT_ID>')
+    dataset = project.version('<PROJECT_VERSION>').download("yolov8")
+    ```
+
+=== "Pascal VOC"
+
+    ```python
+    import roboflow
+
+    roboflow.login()
+
+    rf = roboflow.Roboflow()
+    project = rf.workspace('<WORKSPACE_ID>').project('<PROJECT_ID>')
+    dataset = project.version('<PROJECT_VERSION>').download("voc")
+    ```
+
+## Load Dataset
+
+The Supervision library provides convenient functions to load datasets in various
+formats. If your dataset is already split into train, test, and valid subsets, you can
+load each of those as separate [`sv.DetectionDataset`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset)
+instances.
+
+=== "COCO"
+
+    We can do so using the [`sv.DetectionDataset.from_coco`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.from_coco) to load annotations in [COCO](https://roboflow.com/formats/coco-json) format.
+
+    ```python
+    import supervision as sv
+
+    ds_train = sv.DetectionDataset.from_coco(
+        images_directory_path=f'{dataset.location}/train',
+        annotations_path=f'{dataset.location}/train/_annotations.coco.json',
+    )
+    ds_valid = sv.DetectionDataset.from_coco(
+        images_directory_path=f'{dataset.location}/valid',
+        annotations_path=f'{dataset.location}/valid/_annotations.coco.json',
+    )
+    ds_test = sv.DetectionDataset.from_coco(
+        images_directory_path=f'{dataset.location}/test',
+        annotations_path=f'{dataset.location}/test/_annotations.coco.json',
+    )
+
+    ds_train.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds_train), len(ds_valid), len(ds_test)
+    # 800, 100, 100
+    ```
+
+=== "YOLO"
+
+    We can do so using the [`sv.DetectionDataset.from_yolo`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.from_yolo) to load annotations in [YOLO](https://roboflow.com/formats/yolov8-pytorch-txt) format.
+
+    ```python
+    import supervision as sv
+
+    ds_train = sv.DetectionDataset.from_yolo(
+        images_directory_path=f'{dataset.location}/train/images',
+        annotations_directory_path=f'{dataset.location}/train/labels',
+        data_yaml_path=f'{dataset.location}/data.yaml'
+    )
+    ds_valid = sv.DetectionDataset.from_yolo(
+        images_directory_path=f'{dataset.location}/valid/images',
+        annotations_directory_path=f'{dataset.location}/valid/labels',
+        data_yaml_path=f'{dataset.location}/data.yaml'
+    )
+    ds_test = sv.DetectionDataset.from_yolo(
+        images_directory_path=f'{dataset.location}/test/images',
+        annotations_directory_path=f'{dataset.location}/test/labels',
+        data_yaml_path=f'{dataset.location}/data.yaml'
+    )
+
+    ds_train.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds_train), len(ds_valid), len(ds_test)
+    # 800, 100, 100
+    ```
+
+=== "Pascal VOC"
+
+    We can do so using the [`sv.DetectionDataset.from_pascal_voc`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.from_pascal_voc) to load annotations in [Pascal VOC](https://roboflow.com/formats/pascal-voc-xml) format.
+
+    ```python
+    import supervision as sv
+
+    ds_train = sv.DetectionDataset.from_pascal_voc(
+        images_directory_path=f'{dataset.location}/train/images',
+        annotations_directory_path=f'{dataset.location}/train/labels'
+    )
+    ds_valid = sv.DetectionDataset.from_pascal_voc(
+        images_directory_path=f'{dataset.location}/valid/images',
+        annotations_directory_path=f'{dataset.location}/valid/labels'
+    )
+    ds_test = sv.DetectionDataset.from_pascal_voc(
+        images_directory_path=f'{dataset.location}/test/images',
+        annotations_directory_path=f'{dataset.location}/test/labels'
+    )
+
+    ds_train.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds_train), len(ds_valid), len(ds_test)
+    # 800, 100, 100
+    ```
+
+## Split Dataset
+
+If your dataset is not already split into train, test, and valid subsets, you can
+easily do so using the [`sv.DetectionDataset.split`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.split)
+method. We can split it as follows, ensuring a random shuffle of the data.
+
+```python
+import supervision as sv
+
+ds = sv.DetectionDataset(...)
+
+len(ds)
+# 1000
+
+ds_train, ds = ds.split(split_ratio=0.8, shuffle=True)
+ds_valid, ds_test = ds.split(split_ratio=0.5, shuffle=True)
+
+len(ds_train), len(ds_valid), len(ds_test)
+# 800, 100, 100
+```
+
+## Merge Dataset
+
+If you have multiple datasets that you would like to merge, you can do so using the
+[`sv.DetectionDataset.merge`](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.merge)
+method.
+
+=== "COCO"
+
+    ```{ .py hl_lines="22-28" }
+    import supervision as sv
+
+    ds_train = sv.DetectionDataset.from_coco(
+        images_directory_path=f'{dataset.location}/train',
+        annotations_path=f'{dataset.location}/train/_annotations.coco.json',
+    )
+    ds_valid = sv.DetectionDataset.from_coco(
+        images_directory_path=f'{dataset.location}/valid',
+        annotations_path=f'{dataset.location}/valid/_annotations.coco.json',
+    )
+    ds_test = sv.DetectionDataset.from_coco(
+        images_directory_path=f'{dataset.location}/test',
+        annotations_path=f'{dataset.location}/test/_annotations.coco.json',
+    )
+
+    ds_train.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds_train), len(ds_valid), len(ds_test)
+    # 800, 100, 100
+
+    ds = sv.DetectionDataset.merge([ds_train, ds_valid, ds_test])
+
+    ds.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds)
+    # 1000
+    ```
+
+=== "YOLO"
+
+    ```{ .py hl_lines="25-31" }
+    import supervision as sv
+
+    ds_train = sv.DetectionDataset.from_yolo(
+        images_directory_path=f'{dataset.location}/train/images',
+        annotations_directory_path=f'{dataset.location}/train/labels',
+        data_yaml_path=f'{dataset.location}/data.yaml'
+    )
+    ds_valid = sv.DetectionDataset.from_yolo(
+        images_directory_path=f'{dataset.location}/valid/images',
+        annotations_directory_path=f'{dataset.location}/valid/labels',
+        data_yaml_path=f'{dataset.location}/data.yaml'
+    )
+    ds_test = sv.DetectionDataset.from_yolo(
+        images_directory_path=f'{dataset.location}/test/images',
+        annotations_directory_path=f'{dataset.location}/test/labels',
+        data_yaml_path=f'{dataset.location}/data.yaml'
+    )
+
+    ds_train.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds_train), len(ds_valid), len(ds_test)
+    # 800, 100, 100
+
+    ds = sv.DetectionDataset.merge([ds_train, ds_valid, ds_test])
+
+    ds.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds)
+    # 1000
+    ```
+
+=== "Pascal VOC"
+
+    ```{ .py hl_lines="22-28" }
+    import supervision as sv
+
+    ds_train = sv.DetectionDataset.from_pascal_voc(
+        images_directory_path=f'{dataset.location}/train/images',
+        annotations_directory_path=f'{dataset.location}/train/labels'
+    )
+    ds_valid = sv.DetectionDataset.from_pascal_voc(
+        images_directory_path=f'{dataset.location}/valid/images',
+        annotations_directory_path=f'{dataset.location}/valid/labels'
+    )
+    ds_test = sv.DetectionDataset.from_pascal_voc(
+        images_directory_path=f'{dataset.location}/test/images',
+        annotations_directory_path=f'{dataset.location}/test/labels'
+    )
+
+    ds_train.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds_train), len(ds_valid), len(ds_test)
+    # 800, 100, 100
+
+    ds = sv.DetectionDataset.merge([ds_train, ds_valid, ds_test])
+
+    ds.classes
+    # ['person', 'bicycle', 'car', ...]
+
+    len(ds)
+    # 1000
+    ```
+
+## Iterate over Dataset
+
+There are two ways to loop over a `sv.DetectionDataset`: using a direct
+[for loop](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.__iter__)
+called on the `sv.DetectionDataset` instance or loading `sv.DetectionDataset` entries
+[by index](https://supervision.roboflow.com/latest/datasets/core/#supervision.dataset.core.DetectionDataset.__getitem__).
+
+```python
+import supervision as sv
+
+ds = sv.DetectionDataset(...)
+
+# Option 1
+for image_path, image, annotations in ds:
+    ... # Process each image and its annotations
+
+# Option 2
+for idx in range(len(ds)):
+    image_path, image, annotations = ds[idx]
+    ... # Process the image and annotations at index `idx`
+```
+
+## Visualize Dataset
+
+The Supervision library provides tools for easily visualizing your detection dataset.
+You can create a grid of annotated images to quickly inspect your data and labels.
+First, initialize the [`sv.BoxAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.BoxAnnotator)
+and [`sv.LabelAnnotator`](https://supervision.roboflow.com/latest/detection/annotators/#supervision.annotators.core.LabelAnnotator).
+Then, iterate through a subset of the dataset (e.g., the first 25 images), drawing
+bounding boxes and class labels on each image. Finally, combine the annotated images
+into a grid for display.
+
+```python
+import supervision as sv
+
+ds = sv.DetectionDataset(...)
+
+box_annotator = sv.BoxAnnotator()
+label_annotator = sv.LabelAnnotator()
+
+annotated_images = []
+for i in range(16):
+    _, image, annotations = ds[i]
+
+    labels = [ds.classes[class_id] for class_id in annotations.class_id]
+
+    annotated_image = image.copy()
+    annotated_image = box_annotator.annotate(annotated_image, annotations)
+    annotated_image = label_annotator.annotate(annotated_image, annotations, labels)
+    annotated_images.append(annotated_image)
+
+grid = sv.create_tiles(
+    annotated_images,
+    grid_size=(4, 4),
+    single_tile_size=(400, 400),
+    tile_padding_color=sv.Color.WHITE,
+    tile_margin_color=sv.Color.WHITE
+)
+```
+
+![visualize-dataset](https://media.roboflow.com/supervision-docs/visualize-dataset.png)
+
+## Save Dataset
+
+=== "COCO"
+
+    We can do so using the [`sv.DetectionDataset.as_coco`](https://supervision.roboflow.com/datasets/#supervision.dataset.core.DetectionDataset.as_coco) method to save annotations in [COCO](https://roboflow.com/formats/coco-json) format.
+
+    ```python
+    import supervision as sv
+
+    ds = sv.DetectionDataset(...)
+
+    ds.as_coco(
+        images_directory_path='<IMAGE_DIRECTORY_PATH>',
+        annotations_path='<ANNOTATIONS_PATH>'
+    )
+    ```
+
+=== "YOLO"
+
+    We can do so using the [`sv.DetectionDataset.as_yolo`](https://supervision.roboflow.com/datasets/#supervision.dataset.core.DetectionDataset.as_yolo) method to save annotations in [YOLO](https://roboflow.com/formats/yolov8-pytorch-txt) format.
+
+    ```python
+    import supervision as sv
+
+    ds = sv.DetectionDataset(...)
+
+    ds.as_yolo(
+        images_directory_path='<IMAGE_DIRECTORY_PATH>',
+        annotations_directory_path='<ANNOTATIONS_DIRECTORY_PATH>',
+        data_yaml_path='<DATA_YAML_PATH>'
+    )
+    ```
+
+=== "Pascal VOC"
+
+    We can do so using the [`sv.DetectionDataset.as_pascal_voc`](https://supervision.roboflow.com/datasets/#supervision.dataset.core.DetectionDataset.as_pascal_voc) method to save annotations in [Pascal VOC](https://roboflow.com/formats/pascal-voc-xml) format.
+
+    ```python
+    import supervision as sv
+
+    ds = sv.DetectionDataset(...)
+
+    ds.as_pascal_voc(
+        images_directory_path='<IMAGE_DIRECTORY_PATH>',
+        annotations_directory_path='<ANNOTATIONS_DIRECTORY_PATH>'
+    )
+    ```
+
+## Augment Dataset
+
+In this section, we'll explore using Supervision in combination with Albumentations to
+augment our dataset. Data augmentation is a common technique in computer vision to
+increase the size and diversity of training datasets, leading to improved model
+performance and generalization.
+
+```bash
+pip install augmentation
+```
+
+Albumentations provides a flexible and powerful API for image augmentation. The core of
+the library is the [`Compose`](https://albumentations.ai/docs/api_reference/full_reference/?h=compose#albumentations.core.composition.Compose)
+class, which allows you to chain multiple image transformations together. Each
+transformation is defined using a dedicated class, such as
+[`HorizontalFlip`](https://albumentations.ai/docs/api_reference/full_reference/?h=horizontalflip#albumentations.augmentations.geometric.transforms.HorizontalFlip),
+[`RandomBrightnessContrast`](https://albumentations.ai/docs/api_reference/full_reference/?h=horizontalflip#albumentations.augmentations.transforms.RandomBrightnessContrast),
+or [`Perspective`](https://albumentations.ai/docs/api_reference/full_reference/?h=horizontalflip#albumentations.augmentations.geometric.transforms.Perspective).
+
+```python
+import albumentations as A
+
+augmentation = A.Compose(
+    transforms=[
+        A.Perspective(p=0.1),
+        A.HorizontalFlip(p=0.5),
+        A.RandomBrightnessContrast(p=0.5)
+    ],
+    bbox_params=A.BboxParams(
+        format='pascal_voc',
+        label_fields=['category']
+    ),
+)
+```
+
+The key is to set `format='pascal_voc'`, which corresponds to the
+`[x_min, y_min, x_max, y_max]` bounding box format used in Supervision.
+
+```python
+import numpy as np
+import supervision as sv
+from dataclasses import replace
+
+ds = sv.DetectionDataset(...)
+
+_, original_image, original_annotations = ds[0]
+
+output = augmentation(
+    image=original_image,
+    bboxes=original_annotations.xyxy,
+    category=original_annotations.class_id
+)
+
+augmented_image = output['image']
+augmented_annotations = replace(
+    original_annotations,
+    xyxy=np.array(output['bboxes']),
+    class_id=np.array(output['category'])
+)
+```
+
+![augment-dataset](https://media.roboflow.com/supervision-docs/augment-dataset.png)

docs/how_to/save_detections.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Save Detections

docs/how_to/track_objects.md

@@ -32,7 +32,7 @@ download_assets(VideoAssets.PEOPLE_WALKING)
 First, you'll need to obtain predictions from your object detection or segmentation
 model. In this tutorial, we are using the YOLOv8 model as an example. However,
 Supervision is versatile and compatible with various models. Check this
-[link](latest/how_to/detect_and_annotate/#load-predictions-into-supervision)
+[link](/latest/how_to/detect_and_annotate/#load-predictions-into-supervision)
 for guidance on how to plug in other models.
 
 We will define a `callback` function, which will process each frame of the video

docs/index.md

@@ -146,4 +146,13 @@ You can install `supervision` in a
 
     Master the techniques to selectively filter and focus on objects within a specific zone
 
+-   **Cheatsheet**
+
+    ***
+
+    Access a quick reference guide to the most common `supervision` functions
+
+    [:octicons-arrow-right-24: Cheatsheet](https://roboflow.github.io/cheatsheet-supervision/)
+
+
 </div>

docs/javascripts/cookbooks-card.js

@@ -28,16 +28,15 @@ document.addEventListener("DOMContentLoaded", function () {
     }, {});
 
 
-  async function renderCard(element, elementIndex) {
-    const url = element.getAttribute('data-url');
-    const name = element.getAttribute('data-name');
-    const labels = element.getAttribute('data-labels');
-    const version = element.getAttribute('data-version');
-    const authors = element.getAttribute('data-author');
+    async function renderCard(element, elementIndex) {
+        const name = element.getAttribute('data-name');
+        const labels = element.getAttribute('data-labels');
+        const version = element.getAttribute('data-version');
+        const authors = element.getAttribute('data-author');
 
-    const labelHTML = labels ? labels.split(',').filter(label => label !== '').map((label, index) => {
-        const color = labelToColor[label.trim()];
-        return `
+        const labelHTML = labels ? labels.split(',').filter(label => label !== '').map((label, index) => {
+            const color = labelToColor[label.trim()];
+            return `
             <span
                 class="label non-selectable-text"
                 style="background-color: ${color}"
@@ -45,18 +44,18 @@ document.addEventListener("DOMContentLoaded", function () {
                 ${label.trim()}
             </span>
         `;
-    }).join(' ') : '';
+        }).join(' ') : '';
 
-    const authorArray = authors.split(',');
-    const authorDataArray = await Promise.all(authorArray.map(async (author) => {
-      const response = await fetch(`https://api.github.com/users/${author.trim()}`);
-      return await response.json();
-    }));
+        const authorArray = authors.split(',');
+        const authorDataArray = await Promise.all(authorArray.map(async (author) => {
+            const response = await fetch(`https://api.github.com/users/${author.trim()}`);
+            return await response.json();
+        }));
 
-    let authorAvatarsHTML = authorDataArray.map((authorData, index) => {
-        const marginLeft = index === 0 ? '0' : '-10px';
-        const zIndex = 4 - index;
-        return `
+        let authorAvatarsHTML = authorDataArray.map((authorData, index) => {
+            const marginLeft = index === 0 ? '0' : '-10px';
+            const zIndex = 4 - index;
+            return `
             <div
                 class="author-container"
                 data-login="${authorData.login}-${elementIndex}"
@@ -75,10 +74,10 @@ document.addEventListener("DOMContentLoaded", function () {
                 </a>
             </div>
         `;
-    }).join('');
+        }).join('');
 
-    let authorNamesHTML = authorDataArray.map(
-        authorData => `
+        let authorNamesHTML = authorDataArray.map(
+            authorData => `
             <span
                 class="author-name"
                 data-login="${authorData.login}-${elementIndex}"
@@ -88,54 +87,52 @@ document.addEventListener("DOMContentLoaded", function () {
                 ${authorData.login}
             </a>
         </span>`
-    ).join(',&nbsp;');
+        ).join(',&nbsp;');
 
-    let authorsHTML = `
+        let authorsHTML = `
         <div class="authors">
             ${authorAvatarsHTML}
             <div class="author-names">${authorNamesHTML}</div>
         </div>
-    `;
+        `;
 
-    element.innerText = `
-      <a style="text-decoration: none; color: inherit;" href="${url}">
-        <div style="flex-direction: column; height: 100%; display: flex;
-        font-family: -apple-system,BlinkMacSystemFont,Segoe UI,Helvetica,Arial,sans-serif,Apple Color Emoji,Segoe UI Emoji; background: ${theme.background}; font-size: 14px; line-height: 1.5; color: ${theme.color}">
-          <div style="display: flex; align-items: center;">
-            <span style="font-weight: 700; font-size: 1rem; color: ${theme.linkColor};">
-              ${name}
-            </span>
-          </div>
-          ${authorsHTML}
-          <div style="font-size: 12px; color: ${theme.color}; display: flex; flex: 0; justify-content: space-between">
+        element.innerText = `
+            <div style="flex-direction: column; height: 100%; display: flex;
+            font-family: -apple-system,BlinkMacSystemFont,Segoe UI,Helvetica,Arial,sans-serif,Apple Color Emoji,Segoe UI Emoji; background: ${theme.background}; font-size: 14px; line-height: 1.5; color: ${theme.color}">
             <div style="display: flex; align-items: center;">
-              <img src="/assets/supervision-lenny.png" aria-label="stars" width="20" height="20" role="img" />
-              &nbsp;
-              <span style="margin-left: 4px">${version}</span>
+                <span style="font-weight: 700; font-size: 1rem; color: ${theme.linkColor};">
+                ${name}
+                </span>
             </div>
-            <div style="display: flex; align-items: center; flex-wrap: wrap">
-              ${labelHTML}
+            ${authorsHTML}
+            <div style="font-size: 12px; color: ${theme.color}; display: flex; flex: 0; justify-content: space-between">
+                <div style="display: flex; align-items: center;">
+                <img src="/assets/supervision-lenny.png" aria-label="stars" width="20" height="20" role="img" />
+                &nbsp;
+                <span style="margin-left: 4px">${version}</span>
+                </div>
+                <div style="display: flex; align-items: center; flex-wrap: wrap">
+                ${labelHTML}
+                </div>
             </div>
-          </div>
         </div>
-      </a>
-        `
+        `;
 
-    let sanitizedHTML = DOMPurify.sanitize(element.innerText);
-    element.innerHTML = sanitizedHTML;
+        let sanitizedHTML = DOMPurify.sanitize(element.innerText);
+        element.innerHTML = sanitizedHTML;
 
-    document.querySelectorAll('.author-name').forEach(element => {
-        element.addEventListener('mouseenter', function() {
-            const login = this.getAttribute('data-login');
-            document.querySelector(`.author-container[data-login="${login}"]`).classList.add('hover');
+        document.querySelectorAll('.author-name').forEach(element => {
+            element.addEventListener('mouseenter', function () {
+                const login = this.getAttribute('data-login');
+                document.querySelector(`.author-container[data-login="${login}"]`).classList.add('hover');
+            });
+
+            element.addEventListener('mouseleave', function () {
+                const login = this.getAttribute('data-login');
+                document.querySelector(`.author-container[data-login="${login}"]`).classList.remove('hover');
+            });
         });
-
-        element.addEventListener('mouseleave', function() {
-            const login = this.getAttribute('data-login');
-            document.querySelector(`.author-container[data-login="${login}"]`).classList.remove('hover');
-        });
-    });
-  }
+    }
     repoCards.forEach((element, index) => {
         renderCard(element, index);
     });

docs/keypoint/annotators.md

@@ -13,7 +13,10 @@ status: new
     image = ...
     key_points = sv.KeyPoints(...)
 
-    vertex_annotator = sv.VertexAnnotator(color=sv.Color.GREEN, radius=10)
+    vertex_annotator = sv.VertexAnnotator(
+        color=sv.Color.GREEN,
+        radius=10
+    )
     annotated_frame = vertex_annotator.annotate(
         scene=image.copy(),
         key_points=key_points
@@ -34,7 +37,10 @@ status: new
     image = ...
     key_points = sv.KeyPoints(...)
 
-    edge_annotator = sv.EdgeAnnotator(color=sv.Color.GREEN, thickness=5)
+    edge_annotator = sv.EdgeAnnotator(
+        color=sv.Color.GREEN,
+        thickness=5
+    )
     annotated_frame = edge_annotator.annotate(
         scene=image.copy(),
         key_points=key_points
@@ -47,6 +53,31 @@ status: new
 
     </div>
 
+=== "VertexLabelAnnotator"
+
+    ```python
+    import supervision as sv
+
+    image = ...
+    key_points = sv.KeyPoints(...)
+
+    vertex_label_annotator = sv.VertexLabelAnnotator(
+        color=sv.Color.GREEN,
+        text_color=sv.Color.BLACK,
+        border_radius=5
+    )
+    annotated_frame = vertex_label_annotator.annotate(
+        scene=image.copy(),
+        key_points=key_points
+    )
+    ```
+
+    <div class="result" markdown>
+
+    ![vertex-label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/vertex-label-annotator-example.png){ align=center width="800" }
+
+    </div>
+
 <div class="md-typeset">
   <h2><a href="#supervision.keypoint.annotators.VertexAnnotator">VertexAnnotator</a></h2>
 </div>
@@ -58,3 +89,9 @@ status: new
 </div>
 
 :::supervision.keypoint.annotators.EdgeAnnotator
+
+<div class="md-typeset">
+  <h2><a href="#supervision.keypoint.annotators.VertexLabelAnnotator">VertexLabelAnnotator</a></h2>
+</div>
+
+:::supervision.keypoint.annotators.VertexLabelAnnotator

docs/keypoint/core.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Keypoint Detection

docs/metrics/mean_average_precision.md

@@ -0,0 +1,18 @@
+---
+comments: true
+status: new
+---
+
+# Mean Average Precision
+
+<div class="md-typeset">
+    <h2><a href="#supervision.metrics.mean_average_precision.MeanAveragePrecision">MeanAveragePrecision</a></h2>
+</div>
+
+:::supervision.metrics.mean_average_precision.MeanAveragePrecision
+
+<div class="md-typeset">
+    <h2><a href="#supervision.metrics.mean_average_precision.MeanAveragePrecisionResult">MeanAveragePrecisionResult</a></h2>
+</div>
+
+:::supervision.metrics.mean_average_precision.MeanAveragePrecisionResult

docs/notebooks/annotate-video-with-detections.ipynb

@@ -368,7 +368,7 @@
                 "<br>\n",
                 "\n",
                 "\n",
-                "In the code above we've created a`video_info` variable to pass information about the video to our `VideoSink`. The `VideoSink` is a cool little context manager that allows us to `write_frames()` to a video ouput file. We're also optionally using `tqdm` to display a progress bar with a % complete. We only scratched the surface of all of the customizable Annotators and additional features that Supervision and Inference have to offer. Stay tuned for more cookbooks on how to take advantge of them in your computer vision applications. Happy building! \ud83d\ude80"
+                "In the code above we've created a`video_info` variable to pass information about the video to our `VideoSink`. The `VideoSink` is a cool little context manager that allows us to `write_frames()` to a video output file. We're also optionally using `tqdm` to display a progress bar with a % complete. We only scratched the surface of all of the customizable Annotators and additional features that Supervision and Inference have to offer. Stay tuned for more cookbooks on how to take advantage of them in your computer vision applications. Happy building! \ud83d\ude80"
             ]
         }
     ],

docs/notebooks/count-objects-crossing-the-line.ipynb

@@ -402,7 +402,7 @@
         {
             "cell_type": "markdown",
             "source": [
-                "## Improve Vizualizations"
+                "## Improve Visualizations"
             ],
             "metadata": {
                 "id": "HsPNfuM_3ENR"

docs/notebooks/object-tracking.ipynb

@@ -103,7 +103,7 @@
             "source": [
                 "## Download a Video Asset\n",
                 "\n",
-                "Now that we have our enviornment setup, lets download a video that we can detect objects in. Supervision comes with a great utility to help us hit the ground running. We can use the below snippet to he video is save a video asset in our local directory. It can also be accessed with the variable `path_to_video` for additional application logic."
+                "Now that we have our environment setup, lets download a video that we can detect objects in. Supervision comes with a great utility to help us hit the ground running. We can use the below snippet to he video is save a video asset in our local directory. It can also be accessed with the variable `path_to_video` for additional application logic."
             ]
         },
         {

docs/notebooks/occupancy_analytics.ipynb

@@ -951,7 +951,7 @@
             },
             "source": [
                 "#### Import your data\n",
-                "To load your data back in, upload the saved file to the Colab enviorment and run the code cell."
+                "To load your data back in, upload the saved file to the Colab environment and run the code cell."
             ]
         },
         {

docs/stylesheets/cookbooks-card.css

@@ -19,6 +19,16 @@
     }
 }
 
+.custom-grid a {
+    color: inherit;
+    text-decoration: none;
+    display: block;
+}
+
+.custom-grid a:hover {
+    color: inherit; /* Ensure color does not change on hover */
+}
+
 .repo-card {
     background: radial-gradient(at right top, #A351FB25, #A5F9EA25);
     border-radius: 0.5rem;

docs/theme/cookbooks.html

@@ -11,14 +11,45 @@
       <h1>Supervision Cookbooks</h1>
     </div>
     <div class="custom-grid">
-      <p class="card repo-card" data-url="/develop/notebooks/quickstart" data-name="Supervision Quickstart" data-labels="ANNOTATOR,DETECTION,SAM" data-version="v0.18.0" data-author="SkalskiP,onuralpszr"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/count-objects-crossing-the-line" data-name="Count Objects Crossing the Line" data-labels="ANNOTATORS,LINE ZONE,TRACKING" data-version="v0.18.0" data-author="SkalskiP"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/zero-shot-object-detection-with-yolo-world" data-name="Zero-Shot Object Detection with YOLO-World" data-labels="ANNOTATORS,DETECTION,INFERENCE" data-version="v0.19.0" data-author="SkalskiP"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/download-supervision-assets" data-name="Downloading Supervision Assets" data-labels="ASSETS" data-version="v0.18.0" data-author="nickherrig"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/annotate-video-with-detections" data-name="Annotate Video with Detections" data-labels="INFERENCE,YOLOV8" data-version="v0.18.0" data-author="nickherrig"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/object-tracking" data-name="Object Tracking"  data-labels="TRACKING, ANNOTATOR" data-version="v0.18.0" data-author="nickherrig"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/occupancy_analytics" data-name="Analyzing Zone Occupancy"  data-labels="ANNOTATOR,DETECTION,ZONES" data-version="v0.19.0" data-author="stellasphere"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/evaluating-alignment-of-text-to-image-diffusion-models" data-name="Evaluating Alignment of Text-to-image Diffusion Models"  data-labels="ANNOTATORS,YOLO WORLD" data-version="v0.19.0rc5" data-author="iamhatesz"></p>
+      <a href="/develop/notebooks/quickstart"> <p class="card repo-card" data-name="Supervision Quickstart" data-labels="ANNOTATOR,DETECTION,SAM"
+          data-version="v0.18.0" data-author="SkalskiP,onuralpszr"></p>
+      </a>
+      <a href="/develop/notebooks/count-objects-crossing-the-line">
+        <p class="card repo-card" data-name="Count Objects Crossing the Line"
+          data-labels="ANNOTATORS,LINE ZONE,TRACKING" data-version="v0.18.0" data-author="SkalskiP"></p>
+      </a>
+      <a href="/develop/notebooks/zero-shot-object-detection-with-yolo-world">
+        <p class="card repo-card" data-name="Zero-Shot Object Detection with YOLO-World"
+          data-labels="ANNOTATORS,DETECTION,INFERENCE" data-version="v0.19.0" data-author="SkalskiP"></p>
+      </a>
+      <a href="/develop/notebooks/download-supervision-assets">
+        <p class="card repo-card" data-name="Downloading Supervision Assets" data-labels="ASSETS" data-version="v0.18.0"
+          data-author="nickherrig"></p>
+      </a>
+      <a href="/develop/notebooks/annotate-video-with-detections">
+        <p class="card repo-card" data-name="Annotate Video with Detections" data-labels="INFERENCE,YOLOV8"
+          data-version="v0.18.0" data-author="nickherrig"></p>
+      </a>
+      <a href="/develop/notebooks/object-tracking">
+        <p class="card repo-card" data-name="Object Tracking" data-labels="TRACKING, ANNOTATOR" data-version="v0.18.0"
+          data-author="nickherrig"></p>
+      </a>
+      <a href="/develop/notebooks/occupancy_analytics">
+        <p class="card repo-card" data-name="Analyzing Zone Occupancy" data-labels="ANNOTATOR,DETECTION,ZONES"
+          data-version="v0.19.0" data-author="stellasphere"></p>
+      </a>
+      <a href="/develop/notebooks/evaluating-alignment-of-text-to-image-diffusion-models">
+        <p class="card repo-card" data-name="Evaluating Alignment of Text-to-image Diffusion Models"
+          data-labels="ANNOTATORS,YOLO WORLD" data-version="v0.19.0rc5" data-author="iamhatesz"></p>
+      </a>
+      <a href="/develop/notebooks/serialise-detections-to-csv">
+        <p class="card repo-card" data-name="Serialise Detections to a CSV File"
+          data-labels="DETECTIONS,CSV SINK,INFERENCE" data-version="v0.21.0" data-author="onuralpszr"></p>
+      </a>
+      <a href="/develop/notebooks/serialise-detections-to-json">
+        <p class="card repo-card" data-name="Serialise Detections to a JSON File"
+          data-labels="DETECTIONS,JSON SINK,INFERENCE" data-version="v0.21.0" data-author="onuralpszr"></p>
+      </a>
     </div>
   </div>
 </section>

docs/trackers.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # ByteTrack

docs/utils/draw.md

@@ -41,7 +41,7 @@ comments: true
 :::supervision.draw.utils.draw_image
 
 <div class="md-typeset">
-    <h2><a href="#supervision.draw.utils.calculate_optimal_font_scale">calculate_optimal_font_scale</a></h2>
+    <h2><a href="#supervision.draw.utils.calculate_optimal_text_scale">calculate_optimal_text_scale</a></h2>
 </div>
 
 :::supervision.draw.utils.calculate_optimal_text_scale

docs/utils/image.md

@@ -12,7 +12,7 @@ status: new
 :::supervision.utils.image.crop_image
 
 <div class="md-typeset">
-    <h2><a href="#supervision.utils.image.scale_image">crop_image</a></h2>
+    <h2><a href="#supervision.utils.image.scale_image">scale_image</a></h2>
 </div>
 
 :::supervision.utils.image.scale_image

docs/utils/iterables.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Iterables Utils

docs/utils/video.md

@@ -1,5 +1,6 @@
 ---
 comments: true
+status: new
 ---
 
 # Video Utils

examples/count_people_in_zone/inference_example.py

@@ -11,7 +11,7 @@ from tqdm import tqdm
 
 import supervision as sv
 
-COLORS = sv.ColorPalette.default()
+COLORS = sv.ColorPalette.DEFAULT
 
 
 def load_zones_config(file_path: str) -> List[np.ndarray]:
@@ -38,15 +38,15 @@ def initiate_annotators(
 ) -> Tuple[
     List[sv.PolygonZone], List[sv.PolygonZoneAnnotator], List[sv.BoundingBoxAnnotator]
 ]:
-    line_thickness = sv.calculate_dynamic_line_thickness(resolution_wh=resolution_wh)
-    text_scale = sv.calculate_dynamic_text_scale(resolution_wh=resolution_wh)
+    line_thickness = sv.calculate_optimal_line_thickness(resolution_wh=resolution_wh)
+    text_scale = sv.calculate_optimal_text_scale(resolution_wh=resolution_wh)
 
     zones = []
     zone_annotators = []
     box_annotators = []
 
     for index, polygon in enumerate(polygons):
-        zone = sv.PolygonZone(polygon=polygon, frame_resolution_wh=resolution_wh)
+        zone = sv.PolygonZone(polygon=polygon)
         zone_annotator = sv.PolygonZoneAnnotator(
             zone=zone,
             color=COLORS.by_idx(index),
@@ -75,7 +75,7 @@ def detect(
         frame (np.ndarray): The frame to process, expected to be a NumPy array.
         model (RoboflowInferenceModel): The Inference model used for processing the
             frame.
-        confidence_threshold (float, optional): The confidence threshold for filtering
+        confidence_threshold (float): The confidence threshold for filtering
             detections. Default is 0.5.
 
     Returns:

examples/count_people_in_zone/requirements.txt

@@ -1,5 +1,5 @@
 gdown
-inference
-supervision==0.19.0
+inference==0.9.17
+supervision>=0.20.0
 tqdm
 ultralytics

examples/count_people_in_zone/ultralytics_example.py

@@ -9,7 +9,7 @@ from ultralytics import YOLO
 
 import supervision as sv
 
-COLORS = sv.ColorPalette.default()
+COLORS = sv.ColorPalette.DEFAULT
 
 
 def load_zones_config(file_path: str) -> List[np.ndarray]:
@@ -36,15 +36,15 @@ def initiate_annotators(
 ) -> Tuple[
     List[sv.PolygonZone], List[sv.PolygonZoneAnnotator], List[sv.BoundingBoxAnnotator]
 ]:
-    line_thickness = sv.calculate_dynamic_line_thickness(resolution_wh=resolution_wh)
-    text_scale = sv.calculate_dynamic_text_scale(resolution_wh=resolution_wh)
+    line_thickness = sv.calculate_optimal_line_thickness(resolution_wh=resolution_wh)
+    text_scale = sv.calculate_optimal_text_scale(resolution_wh=resolution_wh)
 
     zones = []
     zone_annotators = []
     box_annotators = []
 
     for index, polygon in enumerate(polygons):
-        zone = sv.PolygonZone(polygon=polygon, frame_resolution_wh=resolution_wh)
+        zone = sv.PolygonZone(polygon=polygon)
         zone_annotator = sv.PolygonZoneAnnotator(
             zone=zone,
             color=COLORS.by_idx(index),
@@ -72,7 +72,7 @@ def detect(
     Args:
         frame (np.ndarray): The frame to process, expected to be a NumPy array.
         model (YOLO): The YOLO model used for processing the frame.
-        confidence_threshold (float, optional): The confidence threshold for filtering
+        confidence_threshold (float): The confidence threshold for filtering
             detections. Default is 0.5.
 
     Returns:

examples/heatmap_and_track/script.py

@@ -20,9 +20,9 @@ def heatmap_and_track(
     iou_threshold: float = 0.5,
     heatmap_alpha: float = 0.5,
     radius: int = 25,
-    track_threshold: float = 0.35,
+    track_activation_threshold: float = 0.35,
     track_seconds: int = 5,
-    match_threshold: float = 0.99,
+    minimum_matching_threshold: float = 0.99,
 ) -> None:
     ### instantiate model
     model = YOLO(source_weights_path)
@@ -47,9 +47,9 @@ def heatmap_and_track(
 
     ### tracker config
     byte_tracker = sv.ByteTrack(
-        track_thresh=track_threshold,
-        track_buffer=track_seconds * fps,
-        match_thresh=match_threshold,
+        track_activation_threshold=track_activation_threshold,
+        lost_track_buffer=track_seconds * fps,
+        minimum_matching_threshold=minimum_matching_threshold,
         frame_rate=fps,
     )
 
@@ -175,7 +175,7 @@ if __name__ == "__main__":
         iou_threshold=args.iou_threshold,
         heatmap_alpha=args.heatmap_alpha,
         radius=args.radius,
-        track_threshold=args.track_threshold,
+        track_activation_threshold=args.track_threshold,
         track_seconds=args.track_seconds,
-        match_threshold=args.match_threshold,
+        minimum_matching_threshold=args.match_threshold,
     )

examples/speed_estimation/inference_example.py

@@ -95,14 +95,14 @@ if __name__ == "__main__":
     model = get_roboflow_model(model_id=args.model_id, api_key=args.roboflow_api_key)
 
     byte_track = sv.ByteTrack(
-        frame_rate=video_info.fps, track_thresh=args.confidence_threshold
+        frame_rate=video_info.fps, track_activation_threshold=args.confidence_threshold
     )
 
-    thickness = sv.calculate_dynamic_line_thickness(
+    thickness = sv.calculate_optimal_line_thickness(
         resolution_wh=video_info.resolution_wh
     )
-    text_scale = sv.calculate_dynamic_text_scale(resolution_wh=video_info.resolution_wh)
-    bounding_box_annotator = sv.BoundingBoxAnnotator(thickness=thickness)
+    text_scale = sv.calculate_optimal_text_scale(resolution_wh=video_info.resolution_wh)
+    box_annotator = sv.BoxAnnotator(thickness=thickness)
     label_annotator = sv.LabelAnnotator(
         text_scale=text_scale,
         text_thickness=thickness,
@@ -116,9 +116,7 @@ if __name__ == "__main__":
 
     frame_generator = sv.get_video_frames_generator(source_path=args.source_video_path)
 
-    polygon_zone = sv.PolygonZone(
-        polygon=SOURCE, frame_resolution_wh=video_info.resolution_wh
-    )
+    polygon_zone = sv.PolygonZone(polygon=SOURCE)
     view_transformer = ViewTransformer(source=SOURCE, target=TARGET)
 
     coordinates = defaultdict(lambda: deque(maxlen=video_info.fps))
@@ -156,7 +154,7 @@ if __name__ == "__main__":
             annotated_frame = trace_annotator.annotate(
                 scene=annotated_frame, detections=detections
             )
-            annotated_frame = bounding_box_annotator.annotate(
+            annotated_frame = box_annotator.annotate(
                 scene=annotated_frame, detections=detections
             )
             annotated_frame = label_annotator.annotate(

examples/speed_estimation/requirements.txt

@@ -1,6 +1,6 @@
-supervision==0.19.0
-tqdm==4.66.1
+supervision>=0.20.0
+tqdm==4.66.3
 requests
 ultralytics==8.0.237
 super-gradients==3.5.0
-inference==0.9.8
+inference==0.9.17

examples/speed_estimation/ultralytics_example.py

@@ -73,14 +73,14 @@ if __name__ == "__main__":
     model = YOLO("yolov8x.pt")
 
     byte_track = sv.ByteTrack(
-        frame_rate=video_info.fps, track_thresh=args.confidence_threshold
+        frame_rate=video_info.fps, track_activation_threshold=args.confidence_threshold
     )
 
-    thickness = sv.calculate_dynamic_line_thickness(
+    thickness = sv.calculate_optimal_line_thickness(
         resolution_wh=video_info.resolution_wh
     )
-    text_scale = sv.calculate_dynamic_text_scale(resolution_wh=video_info.resolution_wh)
-    bounding_box_annotator = sv.BoundingBoxAnnotator(thickness=thickness)
+    text_scale = sv.calculate_optimal_text_scale(resolution_wh=video_info.resolution_wh)
+    box_annotator = sv.BoxAnnotator(thickness=thickness)
     label_annotator = sv.LabelAnnotator(
         text_scale=text_scale,
         text_thickness=thickness,
@@ -94,9 +94,7 @@ if __name__ == "__main__":
 
     frame_generator = sv.get_video_frames_generator(source_path=args.source_video_path)
 
-    polygon_zone = sv.PolygonZone(
-        polygon=SOURCE, frame_resolution_wh=video_info.resolution_wh
-    )
+    polygon_zone = sv.PolygonZone(polygon=SOURCE)
     view_transformer = ViewTransformer(source=SOURCE, target=TARGET)
 
     coordinates = defaultdict(lambda: deque(maxlen=video_info.fps))
@@ -134,7 +132,7 @@ if __name__ == "__main__":
             annotated_frame = trace_annotator.annotate(
                 scene=annotated_frame, detections=detections
             )
-            annotated_frame = bounding_box_annotator.annotate(
+            annotated_frame = box_annotator.annotate(
                 scene=annotated_frame, detections=detections
             )
             annotated_frame = label_annotator.annotate(

examples/speed_estimation/yolo_nas_example.py

@@ -74,14 +74,14 @@ if __name__ == "__main__":
     model = models.get(Models.YOLO_NAS_L, pretrained_weights="coco")
 
     byte_track = sv.ByteTrack(
-        frame_rate=video_info.fps, track_thresh=args.confidence_threshold
+        frame_rate=video_info.fps, track_activation_threshold=args.confidence_threshold
     )
 
-    thickness = sv.calculate_dynamic_line_thickness(
+    thickness = sv.calculate_optimal_line_thickness(
         resolution_wh=video_info.resolution_wh
     )
-    text_scale = sv.calculate_dynamic_text_scale(resolution_wh=video_info.resolution_wh)
-    bounding_box_annotator = sv.BoundingBoxAnnotator(thickness=thickness)
+    text_scale = sv.calculate_optimal_text_scale(resolution_wh=video_info.resolution_wh)
+    box_annotator = sv.BoxAnnotator(thickness=thickness)
     label_annotator = sv.LabelAnnotator(
         text_scale=text_scale,
         text_thickness=thickness,
@@ -95,9 +95,7 @@ if __name__ == "__main__":
 
     frame_generator = sv.get_video_frames_generator(source_path=args.source_video_path)
 
-    polygon_zone = sv.PolygonZone(
-        polygon=SOURCE, frame_resolution_wh=video_info.resolution_wh
-    )
+    polygon_zone = sv.PolygonZone(polygon=SOURCE)
     view_transformer = ViewTransformer(source=SOURCE, target=TARGET)
 
     coordinates = defaultdict(lambda: deque(maxlen=video_info.fps))
@@ -134,7 +132,7 @@ if __name__ == "__main__":
             annotated_frame = trace_annotator.annotate(
                 scene=annotated_frame, detections=detections
             )
-            annotated_frame = bounding_box_annotator.annotate(
+            annotated_frame = box_annotator.annotate(
                 scene=annotated_frame, detections=detections
             )
             annotated_frame = label_annotator.annotate(

examples/time_in_zone/README.md

@@ -103,7 +103,7 @@ python scripts/draw_zones.py \
 ```bash
 python scripts/draw_zones.py \
 --source_path "data/traffic/video.mp4" \
---zone_configuration_path "data/traffic/custom_config.json"
+--zone_configuration_path "data/traffic/config.json"
 ```
 
 https://github.com/roboflow/supervision/assets/26109316/9d514c9e-2a61-418b-ae49-6ac1ad6ae5ac
@@ -157,7 +157,7 @@ Script to run object detection on a video stream using the Roboflow Inference mo
   - `--iou_threshold`: IOU threshold for non-max suppression. Default is `0.7`.
 
 ```bash
-python inference_file_example.py \
+python inference_stream_example.py \
 --zone_configuration_path "data/checkout/config.json" \
 --rtsp_url "rtsp://localhost:8554/live0.stream" \
 --model_id "yolov8x-640" \
@@ -167,7 +167,7 @@ python inference_file_example.py \
 ```
 
 ```bash
-python inference_file_example.py \
+python inference_stream_example.py \
 --zone_configuration_path "data/traffic/config.json" \
 --rtsp_url "rtsp://localhost:8554/live0.stream" \
 --model_id "yolov8x-640" \
@@ -192,7 +192,7 @@ Script to run object detection on a video file using the Ultralytics YOLOv8 mode
   - `--iou_threshold`: IOU threshold for non-max suppression. Default is `0.7`.
 
 ```bash
-python inference_file_example.py \
+python ultralytics_file_example.py \
 --zone_configuration_path "data/checkout/config.json" \
 --source_video_path "data/checkout/video.mp4" \
 --weights "yolov8x.pt" \
@@ -203,7 +203,7 @@ python inference_file_example.py \
 ```
 
 ```bash
-python inference_file_example.py \
+python ultralytics_file_example.py \
 --zone_configuration_path "data/traffic/config.json" \
 --source_video_path "data/traffic/video.mp4" \
 --weights "yolov8x.pt" \
@@ -226,7 +226,7 @@ Script to run object detection on a video stream using the Ultralytics YOLOv8 mo
   - `--iou_threshold`: IOU threshold for non-max suppression. Default is `0.7`.
 
 ```bash
-python inference_file_example.py \
+python ultralytics_stream_example.py \
 --zone_configuration_path "data/checkout/config.json" \
 --rtsp_url "rtsp://localhost:8554/live0.stream" \
 --weights "yolov8x.pt" \
@@ -237,7 +237,7 @@ python inference_file_example.py \
 ```
 
 ```bash
-python inference_file_example.py \
+python ultralytics_stream_example.py \
 --zone_configuration_path "data/traffic/config.json" \
 --rtsp_url "rtsp://localhost:8554/live0.stream" \
 --weights "yolov8x.pt" \

examples/time_in_zone/inference_file_example.py

@@ -29,14 +29,10 @@ def main(
     video_info = sv.VideoInfo.from_video_path(video_path=source_video_path)
     frames_generator = sv.get_video_frames_generator(source_video_path)
 
-    frame = next(frames_generator)
-    resolution_wh = frame.shape[1], frame.shape[0]
-
     polygons = load_zones_config(file_path=zone_configuration_path)
     zones = [
         sv.PolygonZone(
             polygon=polygon,
-            frame_resolution_wh=resolution_wh,
             triggering_anchors=(sv.Position.CENTER,),
         )
         for polygon in polygons

examples/time_in_zone/inference_naive_stream_example.py

@@ -29,14 +29,10 @@ def main(
     frames_generator = get_stream_frames_generator(rtsp_url=rtsp_url)
     fps_monitor = sv.FPSMonitor()
 
-    frame = next(frames_generator)
-    resolution_wh = frame.shape[1], frame.shape[0]
-
     polygons = load_zones_config(file_path=zone_configuration_path)
     zones = [
         sv.PolygonZone(
             polygon=polygon,
-            frame_resolution_wh=resolution_wh,
             triggering_anchors=(sv.Position.CENTER,),
         )
         for polygon in polygons

examples/time_in_zone/inference_stream_example.py

@@ -24,20 +24,15 @@ class CustomSink:
         self.fps_monitor = sv.FPSMonitor()
         self.polygons = load_zones_config(file_path=zone_configuration_path)
         self.timers = [ClockBasedTimer() for _ in self.polygons]
-        self.zones = None
+        self.zones = [
+            sv.PolygonZone(
+                polygon=polygon,
+                triggering_anchors=(sv.Position.CENTER,),
+            )
+            for polygon in self.polygons
+        ]
 
     def on_prediction(self, result: dict, frame: VideoFrame) -> None:
-        if self.zones is None:
-            resolution_wh = frame.image.shape[1], frame.image.shape[0]
-            self.zones = [
-                sv.PolygonZone(
-                    polygon=polygon,
-                    frame_resolution_wh=resolution_wh,
-                    triggering_anchors=(sv.Position.CENTER,),
-                )
-                for polygon in self.polygons
-            ]
-
         self.fps_monitor.tick()
         fps = self.fps_monitor.fps
 

examples/time_in_zone/requirements.txt

@@ -1,5 +1,5 @@
 opencv-python
-supervision
+supervision>=0.20.0
 ultralytics
-inference
+inference==0.9.17
 pytube

examples/time_in_zone/ultralytics_file_example.py

@@ -30,14 +30,10 @@ def main(
     video_info = sv.VideoInfo.from_video_path(video_path=source_video_path)
     frames_generator = sv.get_video_frames_generator(source_video_path)
 
-    frame = next(frames_generator)
-    resolution_wh = frame.shape[1], frame.shape[0]
-
     polygons = load_zones_config(file_path=zone_configuration_path)
     zones = [
         sv.PolygonZone(
             polygon=polygon,
-            frame_resolution_wh=resolution_wh,
             triggering_anchors=(sv.Position.CENTER,),
         )
         for polygon in polygons

examples/time_in_zone/ultralytics_naive_stream_example.py

@@ -30,14 +30,10 @@ def main(
     frames_generator = get_stream_frames_generator(rtsp_url=rtsp_url)
     fps_monitor = sv.FPSMonitor()
 
-    frame = next(frames_generator)
-    resolution_wh = frame.shape[1], frame.shape[0]
-
     polygons = load_zones_config(file_path=zone_configuration_path)
     zones = [
         sv.PolygonZone(
             polygon=polygon,
-            frame_resolution_wh=resolution_wh,
             triggering_anchors=(sv.Position.CENTER,),
         )
         for polygon in polygons

examples/time_in_zone/ultralytics_stream_example.py

@@ -25,20 +25,15 @@ class CustomSink:
         self.fps_monitor = sv.FPSMonitor()
         self.polygons = load_zones_config(file_path=zone_configuration_path)
         self.timers = [ClockBasedTimer() for _ in self.polygons]
-        self.zones = None
+        self.zones = [
+            sv.PolygonZone(
+                polygon=polygon,
+                triggering_anchors=(sv.Position.CENTER,),
+            )
+            for polygon in self.polygons
+        ]
 
     def on_prediction(self, detections: sv.Detections, frame: VideoFrame) -> None:
-        if self.zones is None:
-            resolution_wh = frame.image.shape[1], frame.image.shape[0]
-            self.zones = [
-                sv.PolygonZone(
-                    polygon=polygon,
-                    frame_resolution_wh=resolution_wh,
-                    triggering_anchors=(sv.Position.CENTER,),
-                )
-                for polygon in self.polygons
-            ]
-
         self.fps_monitor.tick()
         fps = self.fps_monitor.fps
 

examples/time_in_zone/utils/timers.py

@@ -22,7 +22,7 @@ class FPSBasedTimer:
         """Initializes the FPSBasedTimer with the specified frames per second rate.
 
         Args:
-            fps (int, optional): The frame rate of the video stream. Defaults to 30.
+            fps (int): The frame rate of the video stream. Defaults to 30.
         """
         self.fps = fps
         self.frame_id = 0

examples/tracking/requirements.txt

@@ -1,4 +1,4 @@
-inference
+inference==0.9.17
 supervision==0.19.0
 tqdm
 ultralytics

examples/traffic_analysis/inference_example.py

@@ -1,6 +1,6 @@
 import argparse
 import os
-from typing import Dict, Iterable, List, Set, Tuple
+from typing import Dict, Iterable, List, Set
 
 import cv2
 import numpy as np
@@ -60,13 +60,11 @@ class DetectionsManager:
 
 def initiate_polygon_zones(
     polygons: List[np.ndarray],
-    frame_resolution_wh: Tuple[int, int],
     triggering_anchors: Iterable[sv.Position] = [sv.Position.CENTER],
 ) -> List[sv.PolygonZone]:
     return [
         sv.PolygonZone(
             polygon=polygon,
-            frame_resolution_wh=frame_resolution_wh,
             triggering_anchors=triggering_anchors,
         )
         for polygon in polygons
@@ -92,14 +90,10 @@ class VideoProcessor:
         self.tracker = sv.ByteTrack()
 
         self.video_info = sv.VideoInfo.from_video_path(source_video_path)
-        self.zones_in = initiate_polygon_zones(
-            ZONE_IN_POLYGONS, self.video_info.resolution_wh, [sv.Position.CENTER]
-        )
-        self.zones_out = initiate_polygon_zones(
-            ZONE_OUT_POLYGONS, self.video_info.resolution_wh, [sv.Position.CENTER]
-        )
+        self.zones_in = initiate_polygon_zones(ZONE_IN_POLYGONS, [sv.Position.CENTER])
+        self.zones_out = initiate_polygon_zones(ZONE_OUT_POLYGONS, [sv.Position.CENTER])
 
-        self.bounding_box_annotator = sv.BoundingBoxAnnotator(color=COLORS)
+        self.box_annotator = sv.BoxAnnotator(color=COLORS)
         self.label_annotator = sv.LabelAnnotator(
             color=COLORS, text_color=sv.Color.BLACK
         )
@@ -140,9 +134,7 @@ class VideoProcessor:
 
         labels = [f"#{tracker_id}" for tracker_id in detections.tracker_id]
         annotated_frame = self.trace_annotator.annotate(annotated_frame, detections)
-        annotated_frame = self.bounding_box_annotator.annotate(
-            annotated_frame, detections
-        )
+        annotated_frame = self.box_annotator.annotate(annotated_frame, detections)
         annotated_frame = self.label_annotator.annotate(
             annotated_frame, detections, labels
         )

examples/traffic_analysis/requirements.txt

@@ -1,5 +1,5 @@
 gdown
-inference
-supervision>=0.19.0
+inference==0.9.17
+supervision>=0.20.0
 tqdm
 ultralytics

examples/traffic_analysis/ultralytics_example.py

@@ -1,5 +1,5 @@
 import argparse
-from typing import Dict, Iterable, List, Set, Tuple
+from typing import Dict, Iterable, List, Set
 
 import cv2
 import numpy as np
@@ -58,13 +58,11 @@ class DetectionsManager:
 
 def initiate_polygon_zones(
     polygons: List[np.ndarray],
-    frame_resolution_wh: Tuple[int, int],
     triggering_anchors: Iterable[sv.Position] = [sv.Position.CENTER],
 ) -> List[sv.PolygonZone]:
     return [
         sv.PolygonZone(
             polygon=polygon,
-            frame_resolution_wh=frame_resolution_wh,
             triggering_anchors=triggering_anchors,
         )
         for polygon in polygons
@@ -89,14 +87,10 @@ class VideoProcessor:
         self.tracker = sv.ByteTrack()
 
         self.video_info = sv.VideoInfo.from_video_path(source_video_path)
-        self.zones_in = initiate_polygon_zones(
-            ZONE_IN_POLYGONS, self.video_info.resolution_wh, [sv.Position.CENTER]
-        )
-        self.zones_out = initiate_polygon_zones(
-            ZONE_OUT_POLYGONS, self.video_info.resolution_wh, [sv.Position.CENTER]
-        )
+        self.zones_in = initiate_polygon_zones(ZONE_IN_POLYGONS, [sv.Position.CENTER])
+        self.zones_out = initiate_polygon_zones(ZONE_OUT_POLYGONS, [sv.Position.CENTER])
 
-        self.bounding_box_annotator = sv.BoundingBoxAnnotator(color=COLORS)
+        self.box_annotator = sv.BoxAnnotator(color=COLORS)
         self.label_annotator = sv.LabelAnnotator(
             color=COLORS, text_color=sv.Color.BLACK
         )
@@ -137,9 +131,7 @@ class VideoProcessor:
 
         labels = [f"#{tracker_id}" for tracker_id in detections.tracker_id]
         annotated_frame = self.trace_annotator.annotate(annotated_frame, detections)
-        annotated_frame = self.bounding_box_annotator.annotate(
-            annotated_frame, detections
-        )
+        annotated_frame = self.box_annotator.annotate(annotated_frame, detections)
         annotated_frame = self.label_annotator.annotate(
             annotated_frame, detections, labels
         )

mkdocs.yml

@@ -33,57 +33,61 @@ extra_css:
   - stylesheets/extra.css
   - stylesheets/cookbooks-card.css
 
-
 nav:
-  - Home: index.md
-  - How to:
+  - Supervision: index.md
+  - Learn:
       - Detect and Annotate: how_to/detect_and_annotate.md
       - Save Detections: how_to/save_detections.md
       - Filter Detections: how_to/filter_detections.md
       - Detect Small Objects: how_to/detect_small_objects.md
-      - Track Objects: how_to/track_objects.md
+      - Track Objects on Video: how_to/track_objects.md
+      - Process Datasets: how_to/process_datasets.md
 
-  - API:
-    - Detection and Segmentation:
-        - Core: detection/core.md
-        - Annotators: detection/annotators.md
-        - Metrics: detection/metrics.md
-        - Utils: detection/utils.md
-    - Keypoint Detection:
-        - Core: keypoint/core.md
-        - Annotators: keypoint/annotators.md
-    - Classification:
-        - Core: classification/core.md
-    - Tools:
-        - Line Zone: detection/tools/line_zone.md
-        - Polygon Zone: detection/tools/polygon_zone.md
-        - Inference Slicer: detection/tools/inference_slicer.md
-        - Detection Smoother: detection/tools/smoother.md
-        - Save Detections: detection/tools/save_detections.md
-    - Trackers: trackers.md
-    - Datasets: datasets.md
-    - Utils:
-        - Video: utils/video.md
-        - Image: utils/image.md
-        - Iterables: utils/iterables.md
-        - Notebook: utils/notebook.md
-        - File: utils/file.md
-        - Draw: utils/draw.md
-        - Geometry: utils/geometry.md
-    - Assets: assets.md
+  - Reference - Code API:
+      - Detection and Segmentation:
+          - Core: detection/core.md
+          - Annotators: detection/annotators.md
+          - Double Detection Filter: detection/double_detection_filter.md
+          - Utils: detection/utils.md
+      - Keypoint Detection:
+          - Core: keypoint/core.md
+          - Annotators: keypoint/annotators.md
+      - Classification:
+          - Core: classification/core.md
+      - Tools:
+          - Line Zone: detection/tools/line_zone.md
+          - Polygon Zone: detection/tools/polygon_zone.md
+          - Inference Slicer: detection/tools/inference_slicer.md
+          - Detection Smoother: detection/tools/smoother.md
+          - Save Detections: detection/tools/save_detections.md
+      - Trackers: trackers.md
+      - Datasets:
+          - Core: datasets/core.md
+          - Utils: datasets/utils.md
+      - Metrics:
+          - mAP: metrics/mean_average_precision.md
+          - Legacy Metrics: detection/metrics.md
+      - Utils:
+          - Video: utils/video.md
+          - Image: utils/image.md
+          - Iterables: utils/iterables.md
+          - Notebook: utils/notebook.md
+          - File: utils/file.md
+          - Draw: utils/draw.md
+          - Geometry: utils/geometry.md
+      - Assets: assets.md
   - Cookbooks: cookbooks.md
+  - Cheatsheet: https://roboflow.github.io/cheatsheet-supervision/
   - Contribute:
-    - Contributing: contributing.md
-    - Code of Conduct: code_of_conduct.md
-    - License: license.md
-  - Changelog:
-        - Changelog: changelog.md
-        - Deprecated: deprecated.md
-
-
+      - Contributing: contributing.md
+      - Code of Conduct: code_of_conduct.md
+      - License: license.md
+  - Release Notes:
+      - Changelog: changelog.md
+      - Deprecated: deprecated.md
 
 theme:
-  name: 'material'
+  name: "material"
   icon:
     edit: material/pencil
   logo: assets/supervision-lenny.png
@@ -100,19 +104,18 @@ theme:
   palette:
     # Palette for light mode
     - scheme: default
-      primary: 'custom'
+      primary: "custom"
       toggle:
         icon: material/brightness-7
         name: Switch to dark mode
 
     # Palette toggle for dark mode
     - scheme: slate
-      primary: 'custom'
+      primary: "custom"
       toggle:
         icon: material/brightness-4
         name: Switch to light mode
 
-
   font:
     text: Roboto
     code: Roboto Mono
@@ -123,10 +126,10 @@ theme:
 plugins:
   - search
   - mkdocs-jupyter:
-        kernel_name: python3
-        execute: false
-        include_source: True
-        include_requirejs: true
+      kernel_name: python3
+      execute: false
+      include_source: True
+      include_requirejs: true
   - mkdocstrings:
       default_handler: python
       handlers:
@@ -150,8 +153,6 @@ plugins:
   - git-revision-date-localized:
       enable_creation_date: true
 
-
-
 markdown_extensions:
   - admonition
   - pymdownx.details
@@ -175,3 +176,12 @@ extra_javascript:
   - "javascripts/init_kapa_widget.js"
   - "javascripts/cookbooks-card.js"
   - "https://cdnjs.cloudflare.com/ajax/libs/dompurify/3.0.8/purify.min.js"
+
+# Messages shown during document build
+# Reference: https://www.mkdocs.org/user-guide/configuration/#validation
+# Values: [warn, info, ignore]
+validation:
+  nav:
+    absolute_links: ignore
+  links:
+    absolute_links: ignore

pyproject.toml

@@ -1,18 +1,26 @@
 [tool.poetry]
 name = "supervision"
-version = "0.20.0"
+version = "0.24.0rc1"
 description = "A set of easy-to-use utils that will come in handy in any Computer Vision project"
 authors = ["Piotr Skalski <piotr.skalski92@gmail.com>"]
 maintainers = ["Piotr Skalski <piotr.skalski92@gmail.com>"]
 readme = "README.md"
 license = "MIT"
-packages = [{include = "supervision"}]
+packages = [{ include = "supervision" }]
 homepage = "https://github.com/roboflow/supervision"
 repository = "https://github.com/roboflow/supervision"
 documentation = "https://github.com/roboflow/supervision/blob/main/README.md"
-keywords = ["machine-learning", "deep-learning", "vision", "ML", "DL", "AI", "Roboflow"]
+keywords = [
+    "machine-learning",
+    "deep-learning",
+    "vision",
+    "ML",
+    "DL",
+    "AI",
+    "Roboflow",
+]
 
-classifiers=[
+classifiers = [
     'Development Status :: 4 - Beta',
     'Intended Audience :: Developers',
     'Intended Audience :: Education',
@@ -27,33 +35,40 @@ classifiers=[
     'Typing :: Typed',
     'Operating System :: Microsoft :: Windows',
     'Operating System :: POSIX :: Linux',
-    'Operating System :: MacOS'
+    'Operating System :: MacOS',
 ]
 
 [tool.poetry.dependencies]
 python = "^3.8"
-numpy = ">=1.21.2"
+numpy = [
+    { version = ">=1.21.2,<1.23.3", python = "<=3.10" },
+    { version = ">=1.23.3", python = ">3.10" },
+]
 scipy = [
     { version = "1.10.0", python = "<3.9" },
-    { version = "^1.10.0", python = ">=3.9" }
+    { version = "^1.10.0", python = ">=3.9" },
 ]
 matplotlib = ">=3.6.0"
 pyyaml = ">=5.3"
 defusedxml = "^0.7.1"
+pillow = ">=9.4"
 opencv-python = { version = ">=4.5.5.64", optional = true }
 opencv-python-headless = ">=4.5.5.64"
-requests = { version = ">=2.26.0,<=2.31.0", optional = true }
-tqdm = { version = ">=4.62.3,<=4.66.2", optional = true }
-pillow = ">=9.4"
+requests = { version = ">=2.26.0,<=2.32.3", optional = true }
+tqdm = { version = ">=4.62.3,<=4.66.5", optional = true }
+# pandas: picked lowest major version that supports Python 3.8
+pandas = { version = ">=2.0.0", optional = true }
+pandas-stubs = { version = ">=2.0.0.230412", optional = true }
 
 [tool.poetry.extras]
 desktop = ["opencv-python"]
-assets = ["requests","tqdm"]
+assets = ["requests", "tqdm"]
+metrics = ["pandas", "pandas-stubs"]
 
 [tool.poetry.group.dev.dependencies]
-twine = ">=4.0.2,<6.0.0"
+twine = "^5.1.1"
 pytest = ">=7.2.2,<9.0.0"
-wheel = ">=0.40,<0.44"
+wheel = ">=0.40,<0.45"
 build = ">=0.10,<1.3"
 ruff = ">=0.1.0"
 mypy = "^1.4.1"
@@ -63,11 +78,14 @@ notebook = ">=6.5.3,<8.0.0"
 ipywidgets = "^8.1.1"
 jupytext = "^1.16.1"
 nbconvert = "^7.14.2"
-
+docutils = [
+    { version = "^0.20.1", python = "<3.9" },
+    { version = "^0.21.1", python = ">=3.9" },
+]
 
 [tool.poetry.group.docs.dependencies]
-mkdocs-material = {extras = ["imaging"], version = "^9.5.5"}
-mkdocstrings = {extras = ["python"], version = ">=0.20,<0.25"}
+mkdocs-material = { extras = ["imaging"], version = "^9.5.5" }
+mkdocstrings = { extras = ["python"], version = ">=0.25.2,<0.27.0" }
 mike = "^2.0.0"
 # For Documentation Development use Python 3.10 or above
 # Use Latest mkdocs-jupyter min 0.24.6 for Jupyter Notebook Theme support
@@ -75,11 +93,15 @@ mkdocs-jupyter = "^0.24.3"
 mkdocs-git-committers-plugin-2 = "^2.2.3"
 mkdocs-git-revision-date-localized-plugin = "^1.2.4"
 
+[tool.poetry.group.typecheck]
+optional = true
 
+[tool.poetry.group.typecheck.dependencies]
+types-pyyaml = "^6.0.12.20240808"
+types-cffi = "^1.16.0.20240331"
+types-requests = "^2.32.0.20240712"
+types-tqdm = "^4.66.0.20240417"
 
-[tool.isort]
-line_length = 88
-profile = "black"
 
 [tool.bandit]
 target = ["test", "supervision"]
@@ -89,7 +111,6 @@ tests = ["B201", "B301", "B318", "B314", "B303", "B413", "B412", "B410"]
 check = true
 imports = ["cv2", "supervision"]
 
-
 [tool.black]
 target-version = ["py38"]
 line-length = 88
@@ -145,10 +166,55 @@ indent-width = 4
 
 [tool.ruff.lint]
 # Enable pycodestyle (`E`) and Pyflakes (`F`) codes by default.
-select = ["E", "F"]
+select = ["E", "F", "I", "A", "Q", "W"]
 ignore = []
 # Allow autofix for all enabled rules (when `--fix`) is provided.
-fixable = ["A", "B", "C", "D", "E", "F", "G", "I", "N", "Q", "S", "T", "W", "ANN", "ARG", "BLE", "COM", "DJ", "DTZ", "EM", "ERA", "EXE", "FBT", "ICN", "INP", "ISC", "NPY", "PD", "PGH", "PIE", "PL", "PT", "PTH", "PYI", "RET", "RSE", "RUF", "SIM", "SLF", "TCH", "TID", "TRY", "UP", "YTT"]
+fixable = [
+    "A",
+    "B",
+    "C",
+    "D",
+    "E",
+    "F",
+    "G",
+    "I",
+    "N",
+    "Q",
+    "S",
+    "T",
+    "W",
+    "ANN",
+    "ARG",
+    "BLE",
+    "COM",
+    "DJ",
+    "DTZ",
+    "EM",
+    "ERA",
+    "EXE",
+    "FBT",
+    "ICN",
+    "INP",
+    "ISC",
+    "NPY",
+    "PD",
+    "PGH",
+    "PIE",
+    "PL",
+    "PT",
+    "PTH",
+    "PYI",
+    "RET",
+    "RSE",
+    "RUF",
+    "SIM",
+    "SLF",
+    "TCH",
+    "TID",
+    "TRY",
+    "UP",
+    "YTT",
+]
 unfixable = []
 # Allow unused variables when underscore-prefixed.
 dummy-variable-rgx = "^(_+|(_+[a-zA-Z0-9_]*[a-zA-Z0-9]+?))$"
@@ -163,12 +229,16 @@ docstring-quotes = "double"
 convention = "google"
 
 [tool.ruff.lint.per-file-ignores]
-"__init__.py" = ["E402","F401"]
+"__init__.py" = ["E402", "F401"]
 
 [tool.ruff.lint.mccabe]
 # Flag errors (`C901`) whenever the complexity level exceeds 5.
 max-complexity = 20
 
+[tool.ruff.lint.isort]
+order-by-type = true
+no-sections = false
+
 [tool.ruff.format]
 # Like Black, use double quotes for strings.
 quote-style = "double"
@@ -186,7 +256,7 @@ line-ending = "auto"
 include-package-data = false
 
 [tool.setuptools.packages.find]
-exclude = ["docs*", "test*","examples*"]
+exclude = ["docs*", "test*", "examples*"]
 
 [build-system]
 requires = ["poetry-core"]

release_process.md

@@ -0,0 +1,24 @@
+# Release Process
+
+This doc outlines how supervision is released into production.
+
+It assumes you already have the code changes, as well as a draft of the release notes.
+
+1. Make sure you have all required changes were merged into `develop`.
+2. Create and merge a PR, merging `develop` into `main`, containing:
+   - A commit that updates the project version in `pyproject.toml`.
+   - All changes made during the release.
+3. Tag the commit with the new supervision version.
+   - make sure to pull from `main` !
+   - Verify that the latest merge commits exists. `git log`.
+   - Run `git tag x.y.z`, with your version
+   - Check with `git log`.
+   - Run `git push origin --tags`
+   - Upon pushing the tag, the [PyPi](https://pypi.org/project/supervision/) should update to the new version. Check this!
+4. Open and merge a PR, merging `main` into `develop`.
+5. Update the docs by running the [Supervision Release Documentation Workflow 📚](https://github.com/roboflow/supervision/actions/workflows/publish-release-docs.yml) workflow from GitHub.
+   - Select the `main` branch from the dropdown.
+6. Create a release on GitHub.
+   - Go to releases
+   - Assign the release notes to the tag created in step 3.
+   - Publish the release.

supervision/__init__.py

@@ -7,8 +7,10 @@ except importlib_metadata.PackageNotFoundError:
     __version__ = "development"
 
 from supervision.annotators.core import (
+    BackgroundOverlayAnnotator,
     BlurAnnotator,
     BoundingBoxAnnotator,
+    BoxAnnotator,
     BoxCornerAnnotator,
     CircleAnnotator,
     ColorAnnotator,
@@ -17,12 +19,14 @@ from supervision.annotators.core import (
     EllipseAnnotator,
     HaloAnnotator,
     HeatMapAnnotator,
+    IconAnnotator,
     LabelAnnotator,
     MaskAnnotator,
     OrientedBoxAnnotator,
     PercentageBarAnnotator,
     PixelateAnnotator,
     PolygonAnnotator,
+    RichLabelAnnotator,
     RoundBoxAnnotator,
     TraceAnnotator,
     TriangleAnnotator,
@@ -34,9 +38,16 @@ from supervision.dataset.core import (
     ClassificationDataset,
     DetectionDataset,
 )
-from supervision.detection.annotate import BoxAnnotator
+from supervision.dataset.utils import mask_to_rle, rle_to_mask
 from supervision.detection.core import Detections
 from supervision.detection.line_zone import LineZone, LineZoneAnnotator
+from supervision.detection.lmm import LMM
+from supervision.detection.overlap_filter import (
+    OverlapFilter,
+    box_non_max_merge,
+    box_non_max_suppression,
+    mask_non_max_suppression,
+)
 from supervision.detection.tools.csv_sink import CSVSink
 from supervision.detection.tools.inference_slicer import InferenceSlicer
 from supervision.detection.tools.json_sink import JSONSink
@@ -44,17 +55,22 @@ from supervision.detection.tools.polygon_zone import PolygonZone, PolygonZoneAnn
 from supervision.detection.tools.smoother import DetectionsSmoother
 from supervision.detection.utils import (
     box_iou_batch,
-    box_non_max_suppression,
     calculate_masks_centroids,
+    clip_boxes,
+    contains_holes,
+    contains_multiple_segments,
     filter_polygons_by_area,
     mask_iou_batch,
-    mask_non_max_suppression,
     mask_to_polygons,
     mask_to_xyxy,
     move_boxes,
+    move_masks,
+    pad_boxes,
     polygon_to_mask,
     polygon_to_xyxy,
     scale_boxes,
+    xcycwh_to_xyxy,
+    xywh_to_xyxy,
 )
 from supervision.draw.color import Color, ColorPalette
 from supervision.draw.utils import (
@@ -69,7 +85,11 @@ from supervision.draw.utils import (
 )
 from supervision.geometry.core import Point, Position, Rect
 from supervision.geometry.utils import get_polygon_center
-from supervision.keypoint.annotators import EdgeAnnotator, VertexAnnotator
+from supervision.keypoint.annotators import (
+    EdgeAnnotator,
+    VertexAnnotator,
+    VertexLabelAnnotator,
+)
 from supervision.keypoint.core import KeyPoints
 from supervision.metrics.detection import ConfusionMatrix, MeanAveragePrecision
 from supervision.tracker.byte_tracker.core import ByteTrack

supervision/annotators/utils.py

@@ -1,5 +1,5 @@
 from enum import Enum
-from typing import Optional, Union
+from typing import Optional, Tuple, Union
 
 import numpy as np
 
@@ -34,14 +34,14 @@ def resolve_color_idx(
 ) -> int:
     if detection_idx >= len(detections):
         raise ValueError(
-            f"Detection index {detection_idx}"
+            f"Detection index {detection_idx} "
             f"is out of bounds for detections of length {len(detections)}"
         )
 
     if isinstance(color_lookup, np.ndarray):
         if len(color_lookup) != len(detections):
             raise ValueError(
-                f"Length of color lookup {len(color_lookup)}"
+                f"Length of color lookup {len(color_lookup)} "
                 f"does not match length of detections {len(detections)}"
             )
         return color_lookup[detection_idx]
@@ -50,19 +50,72 @@ def resolve_color_idx(
     elif color_lookup == ColorLookup.CLASS:
         if detections.class_id is None:
             raise ValueError(
-                "Could not resolve color by class because"
+                "Could not resolve color by class because "
                 "Detections do not have class_id"
             )
         return detections.class_id[detection_idx]
     elif color_lookup == ColorLookup.TRACK:
         if detections.tracker_id is None:
             raise ValueError(
-                "Could not resolve color by track because"
+                "Could not resolve color by track because "
                 "Detections do not have tracker_id"
             )
         return detections.tracker_id[detection_idx]
 
 
+def resolve_text_background_xyxy(
+    center_coordinates: Tuple[int, int],
+    text_wh: Tuple[int, int],
+    position: Position,
+) -> Tuple[int, int, int, int]:
+    center_x, center_y = center_coordinates
+    text_w, text_h = text_wh
+
+    if position == Position.TOP_LEFT:
+        return center_x, center_y - text_h, center_x + text_w, center_y
+    elif position == Position.TOP_RIGHT:
+        return center_x - text_w, center_y - text_h, center_x, center_y
+    elif position == Position.TOP_CENTER:
+        return (
+            center_x - text_w // 2,
+            center_y - text_h,
+            center_x + text_w // 2,
+            center_y,
+        )
+    elif position == Position.CENTER or position == Position.CENTER_OF_MASS:
+        return (
+            center_x - text_w // 2,
+            center_y - text_h // 2,
+            center_x + text_w // 2,
+            center_y + text_h // 2,
+        )
+    elif position == Position.BOTTOM_LEFT:
+        return center_x, center_y, center_x + text_w, center_y + text_h
+    elif position == Position.BOTTOM_RIGHT:
+        return center_x - text_w, center_y, center_x, center_y + text_h
+    elif position == Position.BOTTOM_CENTER:
+        return (
+            center_x - text_w // 2,
+            center_y,
+            center_x + text_w // 2,
+            center_y + text_h,
+        )
+    elif position == Position.CENTER_LEFT:
+        return (
+            center_x - text_w,
+            center_y - text_h // 2,
+            center_x,
+            center_y + text_h // 2,
+        )
+    elif position == Position.CENTER_RIGHT:
+        return (
+            center_x,
+            center_y - text_h // 2,
+            center_x + text_w,
+            center_y + text_h // 2,
+        )
+
+
 def get_color_by_index(color: Union[Color, ColorPalette], idx: int) -> Color:
     if isinstance(color, ColorPalette):
         return color.by_idx(idx)

supervision/dataset/core.py

@@ -3,8 +3,9 @@ from __future__ import annotations
 import os
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
+from itertools import chain
 from pathlib import Path
-from typing import Dict, Iterator, List, Optional, Tuple
+from typing import Dict, Iterator, List, Optional, Tuple, Union
 
 import cv2
 import numpy as np
@@ -31,9 +32,10 @@ from supervision.dataset.utils import (
     train_test_split,
 )
 from supervision.detection.core import Detections
+from supervision.utils.internal import deprecated, warn_deprecated
+from supervision.utils.iterables import find_duplicates
 
 
-@dataclass
 class BaseDataset(ABC):
     @abstractmethod
     def __len__(self) -> int:
@@ -41,35 +43,103 @@ class BaseDataset(ABC):
 
     @abstractmethod
     def split(
-        self, split_ratio=0.8, random_state=None, shuffle: bool = True
+        self,
+        split_ratio: float = 0.8,
+        random_state: Optional[int] = None,
+        shuffle: bool = True,
     ) -> Tuple[BaseDataset, BaseDataset]:
         pass
 
 
-@dataclass
 class DetectionDataset(BaseDataset):
     """
-    Dataclass containing information about object detection dataset.
+    Contains information about a detection dataset. Handles lazy image loading
+    and annotation retrieval, dataset splitting, conversions into multiple
+    formats.
 
     Attributes:
         classes (List[str]): List containing dataset class names.
-        images (Dict[str, np.ndarray]): Dictionary mapping image name to image.
+        images (Union[List[str], Dict[str, np.ndarray]]):
+            Accepts a list of image paths, or dictionaries of loaded cv2 images
+            with paths as keys. If you pass a list of paths, the dataset will
+            lazily load images on demand, which is much more memory-efficient.
         annotations (Dict[str, Detections]): Dictionary mapping
-            image name to annotations.
+            image path to annotations. The dictionary keys match
+            match the keys in `images` or entries in the list of
+            image paths.
     """
 
-    classes: List[str]
-    images: Dict[str, np.ndarray]
-    annotations: Dict[str, Detections]
+    def __init__(
+        self,
+        classes: List[str],
+        images: Union[List[str], Dict[str, np.ndarray]],
+        annotations: Dict[str, Detections],
+    ) -> None:
+        self.classes = classes
+
+        if set(images) != set(annotations):
+            raise ValueError(
+                "The keys of the images and annotations dictionaries must match."
+            )
+        self.annotations = annotations
+
+        # Eliminate duplicates while preserving order
+        self.image_paths = list(dict.fromkeys(images))
+
+        self._images_in_memory: Dict[str, np.ndarray] = {}
+        if isinstance(images, dict):
+            self._images_in_memory = images
+            warn_deprecated(
+                "Passing a `Dict[str, np.ndarray]` into `DetectionDataset` is "
+                "deprecated and will be removed in `supervision-0.26.0`. Use "
+                "a list of paths `List[str]` instead."
+            )
+            # TODO: when supervision-0.26.0 is released, and Dict[str, np.ndarray]
+            #       for images is no longer supported, also simplify the rest of
+            #       the code. E.g. list(images) is no longer needed, and merge can
+            #       be simplified.
+
+    @property
+    @deprecated(
+        "`DetectionDataset.images` property is deprecated and will be removed in "
+        "`supervision-0.26.0`. Iterate with `for path, image, annotation in dataset:` "
+        "instead."
+    )
+    def images(self) -> Dict[str, np.ndarray]:
+        """
+        Load all images to memory and return them as a dictionary.
+
+        !!! warning
+
+            Only use this when you need all images at once.
+            It is much more memory-efficient to initialize dataset with
+            image paths and use `for path, image, annotation in dataset:`.
+        """
+        if self._images_in_memory:
+            return self._images_in_memory
+
+        images = {image_path: cv2.imread(image_path) for image_path in self.image_paths}
+        return images
+
+    def _get_image(self, image_path: str) -> np.ndarray:
+        """Assumes that image is in dataset"""
+        if self._images_in_memory:
+            return self._images_in_memory[image_path]
+        return cv2.imread(image_path)
 
     def __len__(self) -> int:
-        """
-        Return the number of images in the dataset.
+        return len(self._images_in_memory) or len(self.image_paths)
 
-        Returns:
-            int: The number of images.
+    def __getitem__(self, i: int) -> Tuple[str, np.ndarray, Detections]:
         """
-        return len(self.images)
+        Returns:
+            Tuple[str, np.ndarray, Detections]: The image path, image data,
+                and its corresponding annotation at index i.
+        """
+        image_path = self.image_paths[i]
+        image = self._get_image(image_path)
+        annotation = self.annotations[image_path]
+        return image_path, image, annotation
 
     def __iter__(self) -> Iterator[Tuple[str, np.ndarray, Detections]]:
         """
@@ -77,77 +147,190 @@ class DetectionDataset(BaseDataset):
 
         Yields:
             Iterator[Tuple[str, np.ndarray, Detections]]:
-                An iterator that yields tuples containing the image name,
+                An iterator that yields tuples containing the image path,
                 the image data, and its corresponding annotation.
         """
-        for image_name, image in self.images.items():
-            yield image_name, image, self.annotations.get(image_name, None)
+        for i in range(len(self)):
+            image_path, image, annotation = self[i]
+            yield image_path, image, annotation
 
-    def __eq__(self, other):
+    def __eq__(self, other) -> bool:
         if not isinstance(other, DetectionDataset):
             return False
 
         if set(self.classes) != set(other.classes):
             return False
 
-        for key in self.images:
-            if not np.array_equal(self.images[key], other.images[key]):
-                return False
-            if not self.annotations[key] == other.annotations[key]:
+        if self.image_paths != other.image_paths:
+            return False
+
+        if self._images_in_memory or other._images_in_memory:
+            if not np.array_equal(
+                list(self._images_in_memory.values()),
+                list(other._images_in_memory.values()),
+            ):
                 return False
 
+        if self.annotations != other.annotations:
+            return False
+
         return True
 
     def split(
-        self, split_ratio=0.8, random_state=None, shuffle: bool = True
+        self,
+        split_ratio: float = 0.8,
+        random_state: Optional[int] = None,
+        shuffle: bool = True,
     ) -> Tuple[DetectionDataset, DetectionDataset]:
         """
         Splits the dataset into two parts (training and testing)
             using the provided split_ratio.
 
         Args:
-            split_ratio (float, optional): The ratio of the training
+            split_ratio (float): The ratio of the training
                 set to the entire dataset.
-            random_state (int, optional): The seed for the random number generator.
+            random_state (Optional[int]): The seed for the random number generator.
                 This is used for reproducibility.
-            shuffle (bool, optional): Whether to shuffle the data before splitting.
+            shuffle (bool): Whether to shuffle the data before splitting.
 
         Returns:
             Tuple[DetectionDataset, DetectionDataset]: A tuple containing
                 the training and testing datasets.
 
-        Example:
+        Examples:
             ```python
             import supervision as sv
 
             ds = sv.DetectionDataset(...)
-            train_ds, test_ds = ds.split(split_ratio=0.7,
-                                         random_state=42, shuffle=True)
+            train_ds, test_ds = ds.split(split_ratio=0.7, random_state=42, shuffle=True)
             len(train_ds), len(test_ds)
             # (700, 300)
             ```
         """
 
-        image_names = list(self.images.keys())
-        train_names, test_names = train_test_split(
-            data=image_names,
+        train_paths, test_paths = train_test_split(
+            data=self.image_paths,
             train_ratio=split_ratio,
             random_state=random_state,
             shuffle=shuffle,
         )
 
+        train_input: Union[List[str], Dict[str, np.ndarray]]
+        test_input: Union[List[str], Dict[str, np.ndarray]]
+        if self._images_in_memory:
+            train_input = {path: self._images_in_memory[path] for path in train_paths}
+            test_input = {path: self._images_in_memory[path] for path in test_paths}
+        else:
+            train_input = train_paths
+            test_input = test_paths
+        train_annotations = {path: self.annotations[path] for path in train_paths}
+        test_annotations = {path: self.annotations[path] for path in test_paths}
+
         train_dataset = DetectionDataset(
             classes=self.classes,
-            images={name: self.images[name] for name in train_names},
-            annotations={name: self.annotations[name] for name in train_names},
+            images=train_input,
+            annotations=train_annotations,
         )
         test_dataset = DetectionDataset(
             classes=self.classes,
-            images={name: self.images[name] for name in test_names},
-            annotations={name: self.annotations[name] for name in test_names},
+            images=test_input,
+            annotations=test_annotations,
         )
         return train_dataset, test_dataset
 
+    @classmethod
+    def merge(cls, dataset_list: List[DetectionDataset]) -> DetectionDataset:
+        """
+        Merge a list of `DetectionDataset` objects into a single
+            `DetectionDataset` object.
+
+        This method takes a list of `DetectionDataset` objects and combines
+        their respective fields (`classes`, `images`,
+        `annotations`) into a single `DetectionDataset` object.
+
+        Args:
+            dataset_list (List[DetectionDataset]): A list of `DetectionDataset`
+                objects to merge.
+
+        Returns:
+            (DetectionDataset): A single `DetectionDataset` object containing
+            the merged data from the input list.
+
+        Examples:
+            ```python
+            import supervision as sv
+
+            ds_1 = sv.DetectionDataset(...)
+            len(ds_1)
+            # 100
+            ds_1.classes
+            # ['dog', 'person']
+
+            ds_2 = sv.DetectionDataset(...)
+            len(ds_2)
+            # 200
+            ds_2.classes
+            # ['cat']
+
+            ds_merged = sv.DetectionDataset.merge([ds_1, ds_2])
+            len(ds_merged)
+            # 300
+            ds_merged.classes
+            # ['cat', 'dog', 'person']
+            ```
+        """
+
+        def is_in_memory(dataset: DetectionDataset) -> bool:
+            return len(dataset._images_in_memory) > 0 or len(dataset.image_paths) == 0
+
+        def is_lazy(dataset: DetectionDataset) -> bool:
+            return len(dataset._images_in_memory) == 0
+
+        all_in_memory = all([is_in_memory(dataset) for dataset in dataset_list])
+        all_lazy = all([is_lazy(dataset) for dataset in dataset_list])
+        if not all_in_memory and not all_lazy:
+            raise ValueError(
+                "Merging lazy and in-memory DetectionDatasets is not supported."
+            )
+
+        images_in_memory = {}
+        for dataset in dataset_list:
+            images_in_memory.update(dataset._images_in_memory)
+
+        image_paths = list(
+            chain.from_iterable(dataset.image_paths for dataset in dataset_list)
+        )
+        image_paths_unique = list(dict.fromkeys(image_paths))
+        if len(image_paths) != len(image_paths_unique):
+            duplicates = find_duplicates(image_paths)
+            raise ValueError(
+                f"Image paths {duplicates} are not unique across datasets."
+            )
+        image_paths = image_paths_unique
+
+        classes = merge_class_lists(
+            class_lists=[dataset.classes for dataset in dataset_list]
+        )
+
+        annotations = {}
+        for dataset in dataset_list:
+            annotations.update(dataset.annotations)
+        for dataset in dataset_list:
+            class_index_mapping = build_class_index_mapping(
+                source_classes=dataset.classes, target_classes=classes
+            )
+            for image_path in dataset.image_paths:
+                annotations[image_path] = map_detections_class_id(
+                    source_to_target_mapping=class_index_mapping,
+                    detections=annotations[image_path],
+                )
+
+        return cls(
+            classes=classes,
+            images=images_in_memory or image_paths,
+            annotations=annotations,
+        )
+
     def as_pascal_voc(
         self,
         images_directory_path: Optional[str] = None,
@@ -181,25 +364,22 @@ class DetectionDataset(BaseDataset):
         """
         if images_directory_path:
             save_dataset_images(
-                images_directory_path=images_directory_path, images=self.images
+                dataset=self,
+                images_directory_path=images_directory_path,
             )
         if annotations_directory_path:
             Path(annotations_directory_path).mkdir(parents=True, exist_ok=True)
-
-        for image_path, image in self.images.items():
-            detections = self.annotations[image_path]
-
-            if annotations_directory_path:
+            for image_path, image, annotations in self:
                 annotation_name = Path(image_path).stem
                 annotations_path = os.path.join(
                     annotations_directory_path, f"{annotation_name}.xml"
                 )
                 image_name = Path(image_path).name
                 pascal_voc_xml = detections_to_pascal_voc(
-                    detections=detections,
+                    detections=annotations,
                     classes=self.classes,
                     filename=image_name,
-                    image_shape=image.shape,
+                    image_shape=image.shape,  # type: ignore
                     min_image_area_percentage=min_image_area_percentage,
                     max_image_area_percentage=max_image_area_percentage,
                     approximation_percentage=approximation_percentage,
@@ -222,14 +402,14 @@ class DetectionDataset(BaseDataset):
             images_directory_path (str): Path to the directory containing the images.
             annotations_directory_path (str): Path to the directory
                 containing the PASCAL VOC XML annotations.
-            force_masks (bool, optional): If True, forces masks to
+            force_masks (bool): If True, forces masks to
                 be loaded for all annotations, regardless of whether they are present.
 
         Returns:
             DetectionDataset: A DetectionDataset instance containing
                 the loaded images and annotations.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -252,13 +432,15 @@ class DetectionDataset(BaseDataset):
             ```
         """
 
-        classes, images, annotations = load_pascal_voc_annotations(
+        classes, image_paths, annotations = load_pascal_voc_annotations(
             images_directory_path=images_directory_path,
             annotations_directory_path=annotations_directory_path,
             force_masks=force_masks,
         )
 
-        return DetectionDataset(classes=classes, images=images, annotations=annotations)
+        return DetectionDataset(
+            classes=classes, images=image_paths, annotations=annotations
+        )
 
     @classmethod
     def from_yolo(
@@ -267,6 +449,7 @@ class DetectionDataset(BaseDataset):
         annotations_directory_path: str,
         data_yaml_path: str,
         force_masks: bool = False,
+        is_obb: bool = False,
     ) -> DetectionDataset:
         """
         Creates a Dataset instance from YOLO formatted data.
@@ -278,15 +461,18 @@ class DetectionDataset(BaseDataset):
                 containing the YOLO annotation files.
             data_yaml_path (str): The path to the data
                 YAML file containing class information.
-            force_masks (bool, optional): If True, forces
+            force_masks (bool): If True, forces
                 masks to be loaded for all annotations,
                 regardless of whether they are present.
+            is_obb (bool): If True, loads the annotations in OBB format.
+                OBB annotations are defined as `[class_id, x, y, x, y, x, y, x, y]`,
+                where pairs of [x, y] are box corners.
 
         Returns:
             DetectionDataset: A DetectionDataset instance
                 containing the loaded images and annotations.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -308,13 +494,16 @@ class DetectionDataset(BaseDataset):
             # ['dog', 'person']
             ```
         """
-        classes, images, annotations = load_yolo_annotations(
+        classes, image_paths, annotations = load_yolo_annotations(
             images_directory_path=images_directory_path,
             annotations_directory_path=annotations_directory_path,
             data_yaml_path=data_yaml_path,
             force_masks=force_masks,
+            is_obb=is_obb,
+        )
+        return DetectionDataset(
+            classes=classes, images=image_paths, annotations=annotations
         )
-        return DetectionDataset(classes=classes, images=images, annotations=annotations)
 
     def as_yolo(
         self,
@@ -355,13 +544,12 @@ class DetectionDataset(BaseDataset):
         """
         if images_directory_path is not None:
             save_dataset_images(
-                images_directory_path=images_directory_path, images=self.images
+                dataset=self, images_directory_path=images_directory_path
             )
         if annotations_directory_path is not None:
             save_yolo_annotations(
+                dataset=self,
                 annotations_directory_path=annotations_directory_path,
-                images=self.images,
-                annotations=self.annotations,
                 min_image_area_percentage=min_image_area_percentage,
                 max_image_area_percentage=max_image_area_percentage,
                 approximation_percentage=approximation_percentage,
@@ -383,7 +571,7 @@ class DetectionDataset(BaseDataset):
             images_directory_path (str): The path to the
                 directory containing the images.
             annotations_path (str): The path to the json annotation files.
-            force_masks (bool, optional): If True,
+            force_masks (bool): If True,
                 forces masks to be loaded for all annotations,
                 regardless of whether they are present.
 
@@ -391,7 +579,7 @@ class DetectionDataset(BaseDataset):
             DetectionDataset: A DetectionDataset instance containing
                 the loaded images and annotations.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -431,6 +619,20 @@ class DetectionDataset(BaseDataset):
         Exports the dataset to COCO format. This method saves the
         images and their corresponding annotations in COCO format.
 
+        !!! tip
+
+            The format of the mask is determined automatically based on its structure:
+
+            - If a mask contains multiple disconnected components or holes, it will be
+            saved using the Run-Length Encoding (RLE) format for efficient storage and
+            processing.
+            - If a mask consists of a single, contiguous region without any holes, it
+            will be encoded as a polygon, preserving the outline of the object.
+
+            This automatic selection ensures that the masks are stored in the most
+            appropriate and space-efficient format, complying with COCO dataset
+            standards.
+
         Args:
             images_directory_path (Optional[str]): The path to the directory
                 where the images should be saved.
@@ -451,151 +653,195 @@ class DetectionDataset(BaseDataset):
         """
         if images_directory_path is not None:
             save_dataset_images(
-                images_directory_path=images_directory_path, images=self.images
+                dataset=self, images_directory_path=images_directory_path
             )
         if annotations_path is not None:
             save_coco_annotations(
+                dataset=self,
                 annotation_path=annotations_path,
-                images=self.images,
-                annotations=self.annotations,
-                classes=self.classes,
                 min_image_area_percentage=min_image_area_percentage,
                 max_image_area_percentage=max_image_area_percentage,
                 approximation_percentage=approximation_percentage,
             )
 
-    @classmethod
-    def merge(cls, dataset_list: List[DetectionDataset]) -> DetectionDataset:
-        """
-        Merge a list of `DetectionDataset` objects into a single
-            `DetectionDataset` object.
-
-        This method takes a list of `DetectionDataset` objects and combines
-        their respective fields (`classes`, `images`,
-        `annotations`) into a single `DetectionDataset` object.
-
-        Args:
-            dataset_list (List[DetectionDataset]): A list of `DetectionDataset`
-                objects to merge.
-
-        Returns:
-            (DetectionDataset): A single `DetectionDataset` object containing
-            the merged data from the input list.
-
-        Example:
-            ```python
-            import supervision as sv
-
-            ds_1 = sv.DetectionDataset(...)
-            len(ds_1)
-            # 100
-            ds_1.classes
-            # ['dog', 'person']
-
-            ds_2 = sv.DetectionDataset(...)
-            len(ds_2)
-            # 200
-            ds_2.classes
-            # ['cat']
-
-            ds_merged = sv.DetectionDataset.merge([ds_1, ds_2])
-            len(ds_merged)
-            # 300
-            ds_merged.classes
-            # ['cat', 'dog', 'person']
-            ```
-        """
-        merged_images, merged_annotations = {}, {}
-        class_lists = [dataset.classes for dataset in dataset_list]
-        merged_classes = merge_class_lists(class_lists=class_lists)
-
-        for dataset in dataset_list:
-            class_index_mapping = build_class_index_mapping(
-                source_classes=dataset.classes, target_classes=merged_classes
-            )
-            for image_name, image, detections in dataset:
-                if image_name in merged_annotations:
-                    raise ValueError(
-                        f"Image name {image_name} is not unique across datasets."
-                    )
-
-                merged_images[image_name] = image
-                merged_annotations[image_name] = map_detections_class_id(
-                    source_to_target_mapping=class_index_mapping,
-                    detections=detections,
-                )
-
-        return cls(
-            classes=merged_classes, images=merged_images, annotations=merged_annotations
-        )
-
 
 @dataclass
 class ClassificationDataset(BaseDataset):
     """
-    Dataclass containing information about a classification dataset.
+    Contains information about a classification dataset, handles lazy image
+    loading, dataset splitting.
 
     Attributes:
         classes (List[str]): List containing dataset class names.
-        images (Dict[str, np.ndarray]): Dictionary mapping image name to image.
-        annotations (Dict[str, Detections]): Dictionary mapping
+        images (Union[List[str], Dict[str, np.ndarray]]):
+            List of image paths or dictionary mapping image name to image data.
+        annotations (Dict[str, Classifications]): Dictionary mapping
             image name to annotations.
     """
 
-    classes: List[str]
-    images: Dict[str, np.ndarray]
-    annotations: Dict[str, Classifications]
+    def __init__(
+        self,
+        classes: List[str],
+        images: Union[List[str], Dict[str, np.ndarray]],
+        annotations: Dict[str, Classifications],
+    ) -> None:
+        self.classes = classes
+
+        if set(images) != set(annotations):
+            raise ValueError(
+                "The keys of the images and annotations dictionaries must match."
+            )
+        self.annotations = annotations
+
+        # Eliminate duplicates while preserving order
+        self.image_paths = list(dict.fromkeys(images))
+
+        self._images_in_memory: Dict[str, np.ndarray] = {}
+        if isinstance(images, dict):
+            self._images_in_memory = images
+            warn_deprecated(
+                "Passing a `Dict[str, np.ndarray]` into `ClassificationDataset` is "
+                "deprecated and will be removed in a future release. Use "
+                "a list of paths `List[str]` instead."
+            )
+
+    @property
+    @deprecated(
+        "`DetectionDataset.images` property is deprecated and will be removed in "
+        "`supervision-0.26.0`. Iterate with `for path, image, annotation in dataset:` "
+        "instead."
+    )
+    def images(self) -> Dict[str, np.ndarray]:
+        """
+        Load all images to memory and return them as a dictionary.
+
+        !!! warning
+
+            Only use this when you need all images at once.
+            It is much more memory-efficient to initialize dataset with
+            image paths and use `for path, image, annotation in dataset:`.
+        """
+        if self._images_in_memory:
+            return self._images_in_memory
+
+        images = {image_path: cv2.imread(image_path) for image_path in self.image_paths}
+        return images
+
+    def _get_image(self, image_path: str) -> np.ndarray:
+        """Assumes that image is in dataset"""
+        if self._images_in_memory:
+            return self._images_in_memory[image_path]
+        return cv2.imread(image_path)
 
     def __len__(self) -> int:
-        return len(self.images)
+        return len(self._images_in_memory) or len(self.image_paths)
+
+    def __getitem__(self, i: int) -> Tuple[str, np.ndarray, Classifications]:
+        """
+        Returns:
+            Tuple[str, np.ndarray, Classifications]: The image path, image data,
+                and its corresponding annotation at index i.
+        """
+        image_path = self.image_paths[i]
+        image = self._get_image(image_path)
+        annotation = self.annotations[image_path]
+        return image_path, image, annotation
+
+    def __iter__(self) -> Iterator[Tuple[str, np.ndarray, Classifications]]:
+        """
+        Iterate over the images and annotations in the dataset.
+
+        Yields:
+            Iterator[Tuple[str, np.ndarray, Detections]]:
+                An iterator that yields tuples containing the image path,
+                the image data, and its corresponding annotation.
+        """
+        for i in range(len(self)):
+            image_path, image, annotation = self[i]
+            yield image_path, image, annotation
+
+    def __eq__(self, other) -> bool:
+        if not isinstance(other, ClassificationDataset):
+            return False
+
+        if set(self.classes) != set(other.classes):
+            return False
+
+        if self.image_paths != other.image_paths:
+            return False
+
+        if self._images_in_memory or other._images_in_memory:
+            if not np.array_equal(
+                list(self._images_in_memory.values()),
+                list(other._images_in_memory.values()),
+            ):
+                return False
+
+        if self.annotations != other.annotations:
+            return False
+
+        return True
 
     def split(
-        self, split_ratio=0.8, random_state=None, shuffle: bool = True
+        self,
+        split_ratio: float = 0.8,
+        random_state: Optional[int] = None,
+        shuffle: bool = True,
     ) -> Tuple[ClassificationDataset, ClassificationDataset]:
         """
         Splits the dataset into two parts (training and testing)
             using the provided split_ratio.
 
         Args:
-            split_ratio (float, optional): The ratio of the training
+            split_ratio (float): The ratio of the training
                 set to the entire dataset.
-            random_state (int, optional): The seed for the
+            random_state (Optional[int]): The seed for the
                 random number generator. This is used for reproducibility.
-            shuffle (bool, optional): Whether to shuffle the data before splitting.
+            shuffle (bool): Whether to shuffle the data before splitting.
 
         Returns:
             Tuple[ClassificationDataset, ClassificationDataset]: A tuple containing
             the training and testing datasets.
 
-        Example:
+        Examples:
             ```python
             import supervision as sv
 
             cd = sv.ClassificationDataset(...)
-            train_cd,test_cd = cd.split(split_ratio=0.7,
-                                        random_state=42,shuffle=True)
+            train_cd,test_cd = cd.split(split_ratio=0.7, random_state=42,shuffle=True)
             len(train_cd), len(test_cd)
             # (700, 300)
             ```
         """
-        image_names = list(self.images.keys())
-        train_names, test_names = train_test_split(
-            data=image_names,
+        train_paths, test_paths = train_test_split(
+            data=self.image_paths,
             train_ratio=split_ratio,
             random_state=random_state,
             shuffle=shuffle,
         )
 
+        train_input: Union[List[str], Dict[str, np.ndarray]]
+        test_input: Union[List[str], Dict[str, np.ndarray]]
+        if self._images_in_memory:
+            train_input = {path: self._images_in_memory[path] for path in train_paths}
+            test_input = {path: self._images_in_memory[path] for path in test_paths}
+        else:
+            train_input = train_paths
+            test_input = test_paths
+        train_annotations = {path: self.annotations[path] for path in train_paths}
+        test_annotations = {path: self.annotations[path] for path in test_paths}
+
         train_dataset = ClassificationDataset(
             classes=self.classes,
-            images={name: self.images[name] for name in train_names},
-            annotations={name: self.annotations[name] for name in train_names},
+            images=train_input,
+            annotations=train_annotations,
         )
         test_dataset = ClassificationDataset(
             classes=self.classes,
-            images={name: self.images[name] for name in test_names},
-            annotations={name: self.annotations[name] for name in test_names},
+            images=test_input,
+            annotations=test_annotations,
         )
+
         return train_dataset, test_dataset
 
     def as_folder_structure(self, root_directory_path: str) -> None:
@@ -611,18 +857,16 @@ class ClassificationDataset(BaseDataset):
         for class_name in self.classes:
             os.makedirs(os.path.join(root_directory_path, class_name), exist_ok=True)
 
-        for image_path in self.images:
-            classification = self.annotations[image_path]
-            image = self.images[image_path]
-            image_name = Path(image_path).name
+        for image_save_path, image, annotation in self:
+            image_name = Path(image_save_path).name
             class_id = (
-                classification.class_id[0]
-                if classification.confidence is None
-                else classification.get_top_k(1)[0][0]
+                annotation.class_id[0]
+                if annotation.confidence is None
+                else annotation.get_top_k(1)[0][0]
             )
             class_name = self.classes[class_id]
-            image_path = os.path.join(root_directory_path, class_name, image_name)
-            cv2.imwrite(image_path, image)
+            image_save_path = os.path.join(root_directory_path, class_name, image_name)
+            cv2.imwrite(image_save_path, image)
 
     @classmethod
     def from_folder_structure(cls, root_directory_path: str) -> ClassificationDataset:
@@ -635,7 +879,7 @@ class ClassificationDataset(BaseDataset):
         Returns:
             ClassificationDataset: The dataset.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -655,7 +899,7 @@ class ClassificationDataset(BaseDataset):
         classes = os.listdir(root_directory_path)
         classes = sorted(set(classes))
 
-        images = {}
+        image_paths = []
         annotations = {}
 
         for class_name in classes:
@@ -663,13 +907,13 @@ class ClassificationDataset(BaseDataset):
 
             for image in os.listdir(os.path.join(root_directory_path, class_name)):
                 image_path = str(os.path.join(root_directory_path, class_name, image))
-                images[image_path] = cv2.imread(image_path)
+                image_paths.append(image_path)
                 annotations[image_path] = Classifications(
                     class_id=np.array([class_id]),
                 )
 
         return cls(
             classes=classes,
-            images=images,
+            images=image_paths,
             annotations=annotations,
         )

supervision/dataset/formats/coco.py

@@ -1,19 +1,28 @@
 import os
 from datetime import datetime
 from pathlib import Path
-from typing import Dict, List, Tuple
+from typing import TYPE_CHECKING, Dict, List, Tuple
 
-import cv2
 import numpy as np
+import numpy.typing as npt
 
 from supervision.dataset.utils import (
     approximate_mask_with_polygons,
     map_detections_class_id,
+    mask_to_rle,
+    rle_to_mask,
 )
 from supervision.detection.core import Detections
-from supervision.detection.utils import polygon_to_mask
+from supervision.detection.utils import (
+    contains_holes,
+    contains_multiple_segments,
+    polygon_to_mask,
+)
 from supervision.utils.file import read_json_file, save_json_file
 
+if TYPE_CHECKING:
+    from supervision.dataset.core import DetectionDataset
+
 
 def coco_categories_to_classes(coco_categories: List[dict]) -> List[str]:
     return [
@@ -57,13 +66,24 @@ def group_coco_annotations_by_image_id(
     return annotations
 
 
-def _polygons_to_masks(
-    polygons: List[np.ndarray], resolution_wh: Tuple[int, int]
-) -> np.ndarray:
+def coco_annotations_to_masks(
+    image_annotations: List[dict], resolution_wh: Tuple[int, int]
+) -> npt.NDArray[np.bool_]:
     return np.array(
         [
-            polygon_to_mask(polygon=polygon, resolution_wh=resolution_wh)
-            for polygon in polygons
+            rle_to_mask(
+                rle=np.array(image_annotation["segmentation"]["counts"]),
+                resolution_wh=resolution_wh,
+            )
+            if image_annotation["iscrowd"]
+            else polygon_to_mask(
+                polygon=np.reshape(
+                    np.asarray(image_annotation["segmentation"], dtype=np.int32),
+                    (-1, 2),
+                ),
+                resolution_wh=resolution_wh,
+            )
+            for image_annotation in image_annotations
         ],
         dtype=bool,
     )
@@ -83,13 +103,9 @@ def coco_annotations_to_detections(
     xyxy[:, 2:4] += xyxy[:, 0:2]
 
     if with_masks:
-        polygons = [
-            np.reshape(
-                np.asarray(image_annotation["segmentation"], dtype=np.int32), (-1, 2)
-            )
-            for image_annotation in image_annotations
-        ]
-        mask = _polygons_to_masks(polygons=polygons, resolution_wh=resolution_wh)
+        mask = coco_annotations_to_masks(
+            image_annotations=image_annotations, resolution_wh=resolution_wh
+        )
         return Detections(
             class_id=np.asarray(class_ids, dtype=int), xyxy=xyxy, mask=mask
         )
@@ -108,24 +124,35 @@ def detections_to_coco_annotations(
     coco_annotations = []
     for xyxy, mask, _, class_id, _, _ in detections:
         box_width, box_height = xyxy[2] - xyxy[0], xyxy[3] - xyxy[1]
-        polygon = []
+        segmentation = []
+        iscrowd = 0
         if mask is not None:
-            polygon = list(
-                approximate_mask_with_polygons(
-                    mask=mask,
-                    min_image_area_percentage=min_image_area_percentage,
-                    max_image_area_percentage=max_image_area_percentage,
-                    approximation_percentage=approximation_percentage,
-                )[0].flatten()
-            )
+            iscrowd = contains_holes(mask=mask) or contains_multiple_segments(mask=mask)
+
+            if iscrowd:
+                segmentation = {
+                    "counts": mask_to_rle(mask=mask),
+                    "size": list(mask.shape[:2]),
+                }
+            else:
+                segmentation = [
+                    list(
+                        approximate_mask_with_polygons(
+                            mask=mask,
+                            min_image_area_percentage=min_image_area_percentage,
+                            max_image_area_percentage=max_image_area_percentage,
+                            approximation_percentage=approximation_percentage,
+                        )[0].flatten()
+                    )
+                ]
         coco_annotation = {
             "id": annotation_id,
             "image_id": image_id,
             "category_id": int(class_id),
             "bbox": [xyxy[0], xyxy[1], box_width, box_height],
             "area": box_width * box_height,
-            "segmentation": [polygon] if polygon else [],
-            "iscrowd": 0,
+            "segmentation": segmentation,
+            "iscrowd": iscrowd,
         }
         coco_annotations.append(coco_annotation)
         annotation_id += 1
@@ -136,7 +163,7 @@ def load_coco_annotations(
     images_directory_path: str,
     annotations_path: str,
     force_masks: bool = False,
-) -> Tuple[List[str], Dict[str, np.ndarray], Dict[str, Detections]]:
+) -> Tuple[List[str], List[str], Dict[str, Detections]]:
     coco_data = read_json_file(file_path=annotations_path)
     classes = coco_categories_to_classes(coco_categories=coco_data["categories"])
     class_index_mapping = build_coco_class_index_mapping(
@@ -147,7 +174,7 @@ def load_coco_annotations(
         coco_annotations=coco_data["annotations"]
     )
 
-    images = {}
+    images = []
     annotations = {}
 
     for coco_image in coco_images:
@@ -159,7 +186,6 @@ def load_coco_annotations(
         image_annotations = coco_annotations_groups.get(coco_image["id"], [])
         image_path = os.path.join(images_directory_path, image_name)
 
-        image = cv2.imread(image_path)
         annotation = coco_annotations_to_detections(
             image_annotations=image_annotations,
             resolution_wh=(image_width, image_height),
@@ -170,23 +196,20 @@ def load_coco_annotations(
             detections=annotation,
         )
 
-        images[image_path] = image
+        images.append(image_path)
         annotations[image_path] = annotation
 
     return classes, images, annotations
 
 
 def save_coco_annotations(
+    dataset: "DetectionDataset",
     annotation_path: str,
-    images: Dict[str, np.ndarray],
-    annotations: Dict[str, Detections],
-    classes: List[str],
     min_image_area_percentage: float = 0.0,
     max_image_area_percentage: float = 1.0,
     approximation_percentage: float = 0.75,
 ) -> None:
     Path(annotation_path).parent.mkdir(parents=True, exist_ok=True)
-    info = {}
     licenses = [
         {
             "id": 1,
@@ -197,10 +220,10 @@ def save_coco_annotations(
 
     coco_annotations = []
     coco_images = []
-    coco_categories = classes_to_coco_categories(classes=classes)
+    coco_categories = classes_to_coco_categories(classes=dataset.classes)
 
     image_id, annotation_id = 1, 1
-    for image_path, image in images.items():
+    for image_path, image, annotation in dataset:
         image_height, image_width, _ = image.shape
         image_name = f"{Path(image_path).stem}{Path(image_path).suffix}"
         coco_image = {
@@ -213,10 +236,8 @@ def save_coco_annotations(
         }
 
         coco_images.append(coco_image)
-        detections = annotations[image_path]
-
         coco_annotation, annotation_id = detections_to_coco_annotations(
-            detections=detections,
+            detections=annotation,
             image_id=image_id,
             annotation_id=annotation_id,
             min_image_area_percentage=min_image_area_percentage,
@@ -228,7 +249,7 @@ def save_coco_annotations(
         image_id += 1
 
     annotation_dict = {
-        "info": info,
+        "info": {},
         "licenses": licenses,
         "categories": coco_categories,
         "images": coco_images,

supervision/dataset/formats/pascal_voc.py

@@ -5,7 +5,8 @@ from xml.etree.ElementTree import Element, SubElement
 
 import cv2
 import numpy as np
-from defusedxml.ElementTree import fromstring, parse, tostring
+from defusedxml.ElementTree import parse, tostring
+from defusedxml.minidom import parseString
 
 from supervision.dataset.utils import approximate_mask_with_polygons
 from supervision.detection.core import Detections
@@ -129,8 +130,7 @@ def detections_to_pascal_voc(
             annotation.append(next_object)
 
     # Generate XML string
-    xml_string = fromstring(tostring(annotation)).toprettyxml(indent="  ")
-
+    xml_string = parseString(tostring(annotation)).toprettyxml(indent="  ")
     return xml_string
 
 
@@ -138,7 +138,7 @@ def load_pascal_voc_annotations(
     images_directory_path: str,
     annotations_directory_path: str,
     force_masks: bool = False,
-) -> Tuple[List[str], Dict[str, np.ndarray], Dict[str, Detections]]:
+) -> Tuple[List[str], List[str], Dict[str, Detections]]:
     """
     Loads PASCAL VOC XML annotations and returns the image name,
         a Detections instance, and a list of class names.
@@ -147,48 +147,43 @@ def load_pascal_voc_annotations(
         images_directory_path (str): The path to the directory containing the images.
         annotations_directory_path (str): The path to the directory containing the
             PASCAL VOC annotation files.
-        force_masks (bool, optional): If True, forces masks to be loaded for all
+        force_masks (bool): If True, forces masks to be loaded for all
             annotations, regardless of whether they are present.
 
     Returns:
-        Tuple[List[str], Dict[str, np.ndarray], Dict[str, Detections]]: A tuple
-            containing a list of class names,
-            a dictionary with image names as keys and
-            images as values, and a dictionary with image names as
-            keys and corresponding Detections instances as values.
+        Tuple[List[str], List[str], Dict[str, Detections]]: A tuple with a list
+            of class names, a list of paths to images, and a dictionary with image
+            paths as keys and corresponding Detections instances as values.
     """
 
-    image_paths = list_files_with_extensions(
-        directory=images_directory_path, extensions=["jpg", "jpeg", "png"]
-    )
+    image_paths = [
+        str(path)
+        for path in list_files_with_extensions(
+            directory=images_directory_path, extensions=["jpg", "jpeg", "png"]
+        )
+    ]
 
-    classes = []
-    images = {}
+    classes: List[str] = []
     annotations = {}
 
     for image_path in image_paths:
-        image_name = Path(image_path).stem
-        image_path = str(image_path)
-        image = cv2.imread(image_path)
-
-        annotation_path = os.path.join(annotations_directory_path, f"{image_name}.xml")
+        image_stem = Path(image_path).stem
+        annotation_path = os.path.join(annotations_directory_path, f"{image_stem}.xml")
         if not os.path.exists(annotation_path):
-            images[image_path] = image
             annotations[image_path] = Detections.empty()
             continue
 
         tree = parse(annotation_path)
         root = tree.getroot()
 
+        image = cv2.imread(image_path)
         resolution_wh = (image.shape[1], image.shape[0])
         annotation, classes = detections_from_xml_obj(
             root, classes, resolution_wh, force_masks
         )
-
-        images[image_path] = image
         annotations[image_path] = annotation
 
-    return classes, images, annotations
+    return classes, image_paths, annotations
 
 
 def detections_from_xml_obj(

supervision/dataset/formats/yolo.py

@@ -1,10 +1,11 @@
 import os
 from pathlib import Path
-from typing import Dict, List, Optional, Tuple
+from typing import TYPE_CHECKING, Dict, List, Optional, Tuple
 
 import cv2
 import numpy as np
 
+from supervision.config import ORIENTED_BOX_COORDINATES
 from supervision.dataset.utils import approximate_mask_with_polygons
 from supervision.detection.core import Detections
 from supervision.detection.utils import polygon_to_mask, polygon_to_xyxy
@@ -16,6 +17,9 @@ from supervision.utils.file import (
     save_yaml_file,
 )
 
+if TYPE_CHECKING:
+    from supervision.dataset.core import DetectionDataset
+
 
 def _parse_box(values: List[str]) -> np.ndarray:
     x_center, y_center, width, height = values
@@ -70,12 +74,15 @@ def _image_name_to_annotation_name(image_name: str) -> str:
 
 
 def yolo_annotations_to_detections(
-    lines: List[str], resolution_wh: Tuple[int, int], with_masks: bool
+    lines: List[str],
+    resolution_wh: Tuple[int, int],
+    with_masks: bool,
+    is_obb: bool = False,
 ) -> Detections:
     if len(lines) == 0:
         return Detections.empty()
 
-    class_id, relative_xyxy, relative_polygon = [], [], []
+    class_id, relative_xyxy, relative_polygon, relative_xyxyxyxy = [], [], [], []
     w, h = resolution_wh
     for line in lines:
         values = line.split()
@@ -88,21 +95,30 @@ def yolo_annotations_to_detections(
         elif len(values) > 5:
             polygon = _parse_polygon(values=values[1:])
             relative_xyxy.append(polygon_to_xyxy(polygon=polygon))
+            if is_obb:
+                relative_xyxyxyxy.append(np.array(values[1:]))
             if with_masks:
                 relative_polygon.append(polygon)
 
     class_id = np.array(class_id, dtype=int)
     relative_xyxy = np.array(relative_xyxy, dtype=np.float32)
     xyxy = relative_xyxy * np.array([w, h, w, h], dtype=np.float32)
+    data = {}
+
+    if is_obb:
+        relative_xyxyxyxy = np.array(relative_xyxyxyxy, dtype=np.float32)
+        xyxyxyxy = relative_xyxyxyxy.reshape(-1, 4, 2)
+        xyxyxyxy *= np.array([w, h], dtype=np.float32)
+        data[ORIENTED_BOX_COORDINATES] = xyxyxyxy
 
     if not with_masks:
-        return Detections(class_id=class_id, xyxy=xyxy)
+        return Detections(class_id=class_id, xyxy=xyxy, data=data)
 
     polygons = [
         (polygon * np.array(resolution_wh)).astype(int) for polygon in relative_polygon
     ]
     mask = _polygons_to_masks(polygons=polygons, resolution_wh=resolution_wh)
-    return Detections(class_id=class_id, xyxy=xyxy, mask=mask)
+    return Detections(class_id=class_id, xyxy=xyxy, data=data, mask=mask)
 
 
 def load_yolo_annotations(
@@ -110,7 +126,8 @@ def load_yolo_annotations(
     annotations_directory_path: str,
     data_yaml_path: str,
     force_masks: bool = False,
-) -> Tuple[List[str], Dict[str, np.ndarray], Dict[str, Detections]]:
+    is_obb: bool = False,
+) -> Tuple[List[str], List[str], Dict[str, Detections]]:
     """
     Loads YOLO annotations and returns class names, images,
         and their corresponding detections.
@@ -121,34 +138,36 @@ def load_yolo_annotations(
             containing the YOLO annotation files.
         data_yaml_path (str): The path to the data
             YAML file containing class information.
-        force_masks (bool, optional): If True, forces masks to be loaded
+        force_masks (bool): If True, forces masks to be loaded
             for all annotations, regardless of whether they are present.
+        is_obb (bool): If True, loads the annotations in OBB format.
+            OBB annotations are defined as `[class_id, x, y, x, y, x, y, x, y]`,
+            where pairs of [x, y] are box corners.
 
     Returns:
-        Tuple[List[str], Dict[str, np.ndarray], Dict[str, Detections]]:
+        Tuple[List[str], List[str], Dict[str, Detections]]:
             A tuple containing a list of class names, a dictionary with
             image names as keys and images as values, and a dictionary
             with image names as keys and corresponding Detections instances as values.
     """
-    image_paths = list_files_with_extensions(
-        directory=images_directory_path, extensions=["jpg", "jpeg", "png"]
-    )
+    image_paths = [
+        str(path)
+        for path in list_files_with_extensions(
+            directory=images_directory_path, extensions=["jpg", "jpeg", "png"]
+        )
+    ]
 
     classes = _extract_class_names(file_path=data_yaml_path)
-    images = {}
     annotations = {}
 
     for image_path in image_paths:
         image_stem = Path(image_path).stem
-        image_path = str(image_path)
-        image = cv2.imread(image_path)
-
         annotation_path = os.path.join(annotations_directory_path, f"{image_stem}.txt")
         if not os.path.exists(annotation_path):
-            images[image_path] = image
             annotations[image_path] = Detections.empty()
             continue
 
+        image = cv2.imread(image_path)
         lines = read_txt_file(file_path=annotation_path, skip_empty=True)
         h, w, _ = image.shape
         resolution_wh = (w, h)
@@ -156,12 +175,13 @@ def load_yolo_annotations(
         with_masks = _with_mask(lines=lines)
         with_masks = force_masks if force_masks else with_masks
         annotation = yolo_annotations_to_detections(
-            lines=lines, resolution_wh=resolution_wh, with_masks=with_masks
+            lines=lines,
+            resolution_wh=resolution_wh,
+            with_masks=with_masks,
+            is_obb=is_obb,
         )
-
-        images[image_path] = image
         annotations[image_path] = annotation
-    return classes, images, annotations
+    return classes, image_paths, annotations
 
 
 def object_to_yolo(
@@ -195,6 +215,9 @@ def detections_to_yolo_annotations(
 ) -> List[str]:
     annotation = []
     for xyxy, mask, _, class_id, _, _ in detections:
+        if class_id is None:
+            raise ValueError("Class ID is required for YOLO annotations.")
+
         if mask is not None:
             polygons = approximate_mask_with_polygons(
                 mask=mask,
@@ -220,24 +243,22 @@ def detections_to_yolo_annotations(
 
 
 def save_yolo_annotations(
+    dataset: "DetectionDataset",
     annotations_directory_path: str,
-    images: Dict[str, np.ndarray],
-    annotations: Dict[str, Detections],
     min_image_area_percentage: float = 0.0,
     max_image_area_percentage: float = 1.0,
     approximation_percentage: float = 0.75,
 ) -> None:
     Path(annotations_directory_path).mkdir(parents=True, exist_ok=True)
-    for image_path, image in images.items():
-        detections = annotations[image_path]
+    for image_path, image, annotation in dataset:
         image_name = Path(image_path).name
         yolo_annotations_name = _image_name_to_annotation_name(image_name=image_name)
         yolo_annotations_path = os.path.join(
             annotations_directory_path, yolo_annotations_name
         )
         lines = detections_to_yolo_annotations(
-            detections=detections,
-            image_shape=image.shape,
+            detections=annotation,
+            image_shape=image.shape,  # type: ignore
             min_image_area_percentage=min_image_area_percentage,
             max_image_area_percentage=max_image_area_percentage,
             approximation_percentage=approximation_percentage,

supervision/dataset/utils.py

@@ -1,11 +1,13 @@
 import copy
 import os
 import random
+import shutil
 from pathlib import Path
-from typing import Dict, List, Optional, Tuple, TypeVar
+from typing import TYPE_CHECKING, Dict, List, Optional, Tuple, TypeVar, Union
 
 import cv2
 import numpy as np
+import numpy.typing as npt
 
 from supervision.detection.core import Detections
 from supervision.detection.utils import (
@@ -14,6 +16,9 @@ from supervision.detection.utils import (
     mask_to_polygons,
 )
 
+if TYPE_CHECKING:
+    from supervision.dataset.core import DetectionDataset
+
 T = TypeVar("T")
 
 
@@ -58,6 +63,7 @@ def merge_class_lists(class_lists: List[List[str]]) -> List[str]:
 def build_class_index_mapping(
     source_classes: List[str], target_classes: List[str]
 ) -> Dict[int, int]:
+    """Returns the index map of source classes -> target classes."""
     index_mapping = {}
 
     for i, class_name in enumerate(source_classes):
@@ -93,14 +99,16 @@ def map_detections_class_id(
 
 
 def save_dataset_images(
-    images_directory_path: str, images: Dict[str, np.ndarray]
+    dataset: "DetectionDataset", images_directory_path: str
 ) -> None:
     Path(images_directory_path).mkdir(parents=True, exist_ok=True)
-
-    for image_path, image in images.items():
-        image_name = Path(image_path).name
-        target_image_path = os.path.join(images_directory_path, image_name)
-        cv2.imwrite(target_image_path, image)
+    for image_path in dataset.image_paths:
+        final_path = os.path.join(images_directory_path, Path(image_path).name)
+        if image_path in dataset._images_in_memory:
+            image = dataset._images_in_memory[image_path]
+            cv2.imwrite(final_path, image)
+        else:
+            shutil.copyfile(image_path, final_path)
 
 
 def train_test_split(
@@ -129,3 +137,123 @@ def train_test_split(
 
     split_index = int(len(data) * train_ratio)
     return data[:split_index], data[split_index:]
+
+
+def rle_to_mask(
+    rle: Union[npt.NDArray[np.int_], List[int]], resolution_wh: Tuple[int, int]
+) -> npt.NDArray[np.bool_]:
+    """
+    Converts run-length encoding (RLE) to a binary mask.
+
+    Args:
+        rle (Union[npt.NDArray[np.int_], List[int]]): The 1D RLE array, the format
+            used in the COCO dataset (column-wise encoding, values of an array with
+            even indices represent the number of pixels assigned as background,
+            values of an array with odd indices represent the number of pixels
+            assigned as foreground object).
+        resolution_wh (Tuple[int, int]): The width (w) and height (h)
+            of the desired binary mask.
+
+    Returns:
+        The generated 2D Boolean mask of shape `(h, w)`, where the foreground object is
+            marked with `True`'s and the rest is filled with `False`'s.
+
+    Raises:
+        AssertionError: If the sum of pixels encoded in RLE differs from the
+            number of pixels in the expected mask (computed based on resolution_wh).
+
+    Examples:
+        ```python
+        import supervision as sv
+
+        sv.rle_to_mask([5, 2, 2, 2, 5], (4, 4))
+        # array([
+        #     [False, False, False, False],
+        #     [False, True,  True,  False],
+        #     [False, True,  True,  False],
+        #     [False, False, False, False],
+        # ])
+        ```
+    """
+    if isinstance(rle, list):
+        rle = np.array(rle, dtype=int)
+
+    width, height = resolution_wh
+
+    assert width * height == np.sum(rle), (
+        "the sum of the number of pixels in the RLE must be the same "
+        "as the number of pixels in the expected mask"
+    )
+
+    zero_one_values = np.zeros(shape=(rle.size, 1), dtype=np.uint8)
+    zero_one_values[1::2] = 1
+
+    decoded_rle = np.repeat(zero_one_values, rle, axis=0)
+    decoded_rle = np.append(
+        decoded_rle, np.zeros(width * height - len(decoded_rle), dtype=np.uint8)
+    )
+    return decoded_rle.reshape((height, width), order="F")
+
+
+def mask_to_rle(mask: npt.NDArray[np.bool_]) -> List[int]:
+    """
+    Converts a binary mask into a run-length encoding (RLE).
+
+    Args:
+        mask (npt.NDArray[np.bool_]): 2D binary mask where `True` indicates foreground
+            object and `False` indicates background.
+
+    Returns:
+        The run-length encoded mask. Values of a list with even indices
+            represent the number of pixels assigned as background (`False`), values
+            of a list with odd indices represent the number of pixels assigned
+            as foreground object (`True`).
+
+    Raises:
+        AssertionError: If input mask is not 2D or is empty.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([
+            [True, True, True, True],
+            [True, True, True, True],
+            [True, True, True, True],
+            [True, True, True, True],
+        ])
+        sv.mask_to_rle(mask)
+        # [0, 16]
+
+        mask = np.array([
+            [False, False, False, False],
+            [False, True,  True,  False],
+            [False, True,  True,  False],
+            [False, False, False, False],
+        ])
+        sv.mask_to_rle(mask)
+        # [5, 2, 2, 2, 5]
+        ```
+
+    ![mask_to_rle](https://media.roboflow.com/supervision-docs/mask-to-rle.png){ align=center width="800" }
+    """  # noqa E501 // docs
+    assert mask.ndim == 2, "Input mask must be 2D"
+    assert mask.size != 0, "Input mask cannot be empty"
+
+    on_value_change_indices = np.where(
+        mask.ravel(order="F") != np.roll(mask.ravel(order="F"), 1)
+    )[0]
+
+    on_value_change_indices = np.append(on_value_change_indices, mask.size)
+    # need to add 0 at the beginning when the same value is in the first and
+    # last element of the flattened mask
+    if on_value_change_indices[0] != 0:
+        on_value_change_indices = np.insert(on_value_change_indices, 0, 0)
+
+    rle = np.diff(on_value_change_indices)
+
+    if mask[0][0] == 1:
+        rle = np.insert(rle, 0, 0)
+
+    return list(rle)

supervision/detection/annotate.py

@@ -1,155 +0,0 @@
-from typing import List, Optional, Union
-
-import cv2
-
-from supervision.annotators.base import ImageType
-from supervision.detection.core import Detections
-from supervision.draw.color import Color, ColorPalette
-from supervision.utils.conversion import convert_for_annotation_method
-from supervision.utils.internal import deprecated
-
-
-class BoxAnnotator:
-    """
-    A class for drawing bounding boxes on an image using detections provided.
-
-    Attributes:
-        color (Union[Color, ColorPalette]): The color to draw the bounding box,
-            can be a single color or a color palette
-        thickness (int): The thickness of the bounding box lines, default is 2
-        text_color (Color): The color of the text on the bounding box, default is white
-        text_scale (float): The scale of the text on the bounding box, default is 0.5
-        text_thickness (int): The thickness of the text on the bounding box,
-            default is 1
-        text_padding (int): The padding around the text on the bounding box,
-            default is 5
-
-    """
-
-    def __init__(
-        self,
-        color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
-        thickness: int = 2,
-        text_color: Color = Color.BLACK,
-        text_scale: float = 0.5,
-        text_thickness: int = 1,
-        text_padding: int = 10,
-    ):
-        self.color: Union[Color, ColorPalette] = color
-        self.thickness: int = thickness
-        self.text_color: Color = text_color
-        self.text_scale: float = text_scale
-        self.text_thickness: int = text_thickness
-        self.text_padding: int = text_padding
-
-    @deprecated(
-        "`BoxAnnotator` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `BoundingBoxAnnotator` and `LabelAnnotator` instead"
-    )
-    @convert_for_annotation_method
-    def annotate(
-        self,
-        scene: ImageType,
-        detections: Detections,
-        labels: Optional[List[str]] = None,
-        skip_label: bool = False,
-    ) -> ImageType:
-        """
-        Draws bounding boxes on the frame using the detections provided.
-
-        Args:
-            scene (ImageType): The image on which the bounding boxes will be drawn.
-                `ImageType` is a flexible type, accepting either `numpy.ndarray`
-                or `PIL.Image.Image`.
-            detections (Detections): The detections for which the
-                bounding boxes will be drawn
-            labels (Optional[List[str]]): An optional list of labels
-                corresponding to each detection. If `labels` are not provided,
-                corresponding `class_id` will be used as label.
-            skip_label (bool): Is set to `True`, skips bounding box label annotation.
-        Returns:
-            ImageType: The image with the bounding boxes drawn on it, matching the
-                type of `scene` (`numpy.ndarray` or `PIL.Image.Image`)
-
-        Example:
-            ```python
-            import supervision as sv
-
-            classes = ['person', ...]
-            image = ...
-            detections = sv.Detections(...)
-
-            box_annotator = sv.BoxAnnotator()
-            labels = [
-                f"{classes[class_id]} {confidence:0.2f}"
-                for _, _, confidence, class_id, _ in detections
-            ]
-            annotated_frame = box_annotator.annotate(
-                scene=image.copy(),
-                detections=detections,
-                labels=labels
-            )
-            ```
-        """
-        font = cv2.FONT_HERSHEY_SIMPLEX
-        for i in range(len(detections)):
-            x1, y1, x2, y2 = detections.xyxy[i].astype(int)
-            class_id = (
-                detections.class_id[i] if detections.class_id is not None else None
-            )
-            idx = class_id if class_id is not None else i
-            color = (
-                self.color.by_idx(idx)
-                if isinstance(self.color, ColorPalette)
-                else self.color
-            )
-            cv2.rectangle(
-                img=scene,
-                pt1=(x1, y1),
-                pt2=(x2, y2),
-                color=color.as_bgr(),
-                thickness=self.thickness,
-            )
-            if skip_label:
-                continue
-
-            text = (
-                f"{class_id}"
-                if (labels is None or len(detections) != len(labels))
-                else labels[i]
-            )
-
-            text_width, text_height = cv2.getTextSize(
-                text=text,
-                fontFace=font,
-                fontScale=self.text_scale,
-                thickness=self.text_thickness,
-            )[0]
-
-            text_x = x1 + self.text_padding
-            text_y = y1 - self.text_padding
-
-            text_background_x1 = x1
-            text_background_y1 = y1 - 2 * self.text_padding - text_height
-
-            text_background_x2 = x1 + 2 * self.text_padding + text_width
-            text_background_y2 = y1
-
-            cv2.rectangle(
-                img=scene,
-                pt1=(text_background_x1, text_background_y1),
-                pt2=(text_background_x2, text_background_y2),
-                color=color.as_bgr(),
-                thickness=cv2.FILLED,
-            )
-            cv2.putText(
-                img=scene,
-                text=text,
-                org=(text_x, text_y),
-                fontFace=font,
-                fontScale=self.text_scale,
-                color=self.text_color.as_rgb(),
-                thickness=self.text_thickness,
-                lineType=cv2.LINE_AA,
-            )
-        return scene

supervision/detection/core.py

@@ -7,20 +7,35 @@ from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
 import numpy as np
 
 from supervision.config import CLASS_NAME_DATA_FIELD, ORIENTED_BOX_COORDINATES
-from supervision.detection.utils import (
+from supervision.detection.lmm import (
+    LMM,
+    from_florence_2,
+    from_paligemma,
+    validate_lmm_parameters,
+)
+from supervision.detection.overlap_filter import (
+    box_non_max_merge,
     box_non_max_suppression,
+    mask_non_max_suppression,
+)
+from supervision.detection.tools.transformers import (
+    process_transformers_detection_result,
+    process_transformers_v4_segmentation_result,
+    process_transformers_v5_segmentation_result,
+)
+from supervision.detection.utils import (
+    box_iou_batch,
     calculate_masks_centroids,
     extract_ultralytics_masks,
     get_data_item,
     is_data_equal,
-    mask_non_max_suppression,
     mask_to_xyxy,
     merge_data,
     process_roboflow_result,
     xywh_to_xyxy,
 )
 from supervision.geometry.core import Position
-from supervision.utils.internal import deprecated
+from supervision.utils.internal import get_instance_variables
 from supervision.validators import validate_detections_fields
 
 
@@ -234,13 +249,9 @@ class Detections:
             results = model(image)[0]
             detections = sv.Detections.from_ultralytics(results)
             ```
-
-        !!! tip
-
-            Class names values can be accessed using `detections["class_name"]`.
         """  # noqa: E501 // docs
 
-        if "obb" in ultralytics_results and ultralytics_results.obb is not None:
+        if hasattr(ultralytics_results, "obb") and ultralytics_results.obb is not None:
             class_id = ultralytics_results.obb.cls.cpu().numpy().astype(int)
             class_names = np.array([ultralytics_results.names[i] for i in class_id])
             oriented_box_coordinates = ultralytics_results.obb.xyxyxyxy.cpu().numpy()
@@ -257,6 +268,14 @@ class Detections:
                 },
             )
 
+        if hasattr(ultralytics_results, "boxes") and ultralytics_results.boxes is None:
+            masks = extract_ultralytics_masks(ultralytics_results)
+            return cls(
+                xyxy=mask_to_xyxy(masks),
+                mask=masks,
+                class_id=np.arange(len(ultralytics_results)),
+            )
+
         class_id = ultralytics_results.boxes.cls.cpu().numpy().astype(int)
         class_names = np.array([ultralytics_results.names[i] for i in class_id])
         return cls(
@@ -418,6 +437,9 @@ class Detections:
             xyxy=mmdet_results.pred_instances.bboxes.cpu().numpy(),
             confidence=mmdet_results.pred_instances.scores.cpu().numpy(),
             class_id=mmdet_results.pred_instances.labels.cpu().numpy().astype(int),
+            mask=mmdet_results.pred_instances.masks.cpu().numpy()
+            if "masks" in mmdet_results.pred_instances
+            else None,
         )
 
     @classmethod
@@ -425,15 +447,19 @@ class Detections:
         cls, transformers_results: dict, id2label: Optional[Dict[int, str]] = None
     ) -> Detections:
         """
-        Creates a Detections instance from object detection or segmentation
+        Creates a Detections instance from object detection or panoptic, semantic
+        and instance segmentation
         [Transformer](https://github.com/huggingface/transformers) inference result.
 
         Args:
-            transformers_results (dict): The output of Transformers model inference. A
-                dictionary containing the `scores`, `labels`, `boxes` and `masks` keys.
+            transformers_results (Union[dict, torch.Tensor]):  Inference results from
+                your Transformers model. This can be either a dictionary containing
+                valuable outputs like `scores`, `labels`, `boxes`, `masks`,
+                `segments_info`, and `segmentation`, or a `torch.Tensor` holding a
+                segmentation map where values represent class IDs.
             id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to
-                class names. If provided, the resulting Detections object will contain
-                `class_name` data field with the class names.
+                labels, typically part of the `transformers` model configuration. If
+                provided, the resulting dictionary will include class names.
 
         Returns:
             Detections: A new Detections object.
@@ -464,36 +490,28 @@ class Detections:
                 id2label=model.config.id2label
             )
             ```
-
-        !!! tip
-
-            Class names values can be accessed using `detections["class_name"]`.
         """  # noqa: E501 // docs
 
-        class_ids = transformers_results["labels"].cpu().detach().numpy().astype(int)
-        data = {}
-        if id2label is not None:
-            class_names = np.array([id2label[class_id] for class_id in class_ids])
-            data[CLASS_NAME_DATA_FIELD] = class_names
+        if (
+            transformers_results.__class__.__name__ == "Tensor"
+            or "segmentation" in transformers_results
+        ):
+            return cls(
+                **process_transformers_v5_segmentation_result(
+                    transformers_results, id2label
+                )
+            )
+
+        if "masks" in transformers_results or "png_string" in transformers_results:
+            return cls(
+                **process_transformers_v4_segmentation_result(
+                    transformers_results, id2label
+                )
+            )
+
         if "boxes" in transformers_results:
             return cls(
-                xyxy=transformers_results["boxes"].cpu().detach().numpy(),
-                confidence=transformers_results["scores"].cpu().detach().numpy(),
-                class_id=class_ids,
-                data=data,
-            )
-        elif "masks" in transformers_results:
-            masks = transformers_results["masks"].cpu().detach().numpy().astype(bool)
-            return cls(
-                xyxy=mask_to_xyxy(masks),
-                mask=masks,
-                confidence=transformers_results["scores"].cpu().detach().numpy(),
-                class_id=class_ids,
-                data=data,
-            )
-        else:
-            raise NotImplementedError(
-                "Only object detection and semantic segmentation results are supported."
+                **process_transformers_detection_result(transformers_results, id2label)
             )
 
     @classmethod
@@ -532,6 +550,9 @@ class Detections:
         return cls(
             xyxy=detectron2_results["instances"].pred_boxes.tensor.cpu().numpy(),
             confidence=detectron2_results["instances"].scores.cpu().numpy(),
+            mask=detectron2_results["instances"].pred_masks.cpu().numpy()
+            if hasattr(detectron2_results["instances"], "pred_masks")
+            else None,
             class_id=detectron2_results["instances"]
             .pred_classes.cpu()
             .numpy()
@@ -567,10 +588,6 @@ class Detections:
             result = model.infer(image)[0]
             detections = sv.Detections.from_inference(result)
             ```
-
-        !!! tip
-
-            Class names values can be accessed using `detections["class_name"]`.
         """
         with suppress(AttributeError):
             roboflow_result = roboflow_result.dict(exclude_none=True, by_alias=True)
@@ -592,45 +609,6 @@ class Detections:
             data=data,
         )
 
-    @classmethod
-    @deprecated(
-        "`Detections.from_roboflow` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Detections.from_inference` instead."
-    )
-    def from_roboflow(cls, roboflow_result: Union[dict, Any]) -> Detections:
-        """
-        !!! failure "Deprecated"
-
-            `Detections.from_roboflow` is deprecated and will be removed in
-            `supervision-0.22.0`. Use `Detections.from_inference` instead.
-
-        Create a Detections object from the [Roboflow](https://roboflow.com/)
-            API inference result or the [Inference](https://inference.roboflow.com/)
-            package results.
-
-        Args:
-            roboflow_result (dict): The result from the
-                Roboflow API containing predictions.
-
-        Returns:
-            (Detections): A Detections object containing the bounding boxes, class IDs,
-                and confidences of the predictions.
-
-        Example:
-            ```python
-            import cv2
-            import supervision as sv
-            from inference import get_model
-
-            image = cv2.imread(<SOURCE_IMAGE_PATH>)
-            model = get_model(model_id="yolov8s-640")
-
-            result = model.infer(image)[0]
-            detections = sv.Detections.from_roboflow(result)
-            ```
-        """
-        return cls.from_inference(roboflow_result)
-
     @classmethod
     def from_sam(cls, sam_result: List[dict]) -> Detections:
         """
@@ -670,7 +648,7 @@ class Detections:
         if np.asarray(xywh).shape[0] == 0:
             return cls.empty()
 
-        xyxy = xywh_to_xyxy(boxes_xywh=xywh)
+        xyxy = xywh_to_xyxy(xywh=xywh)
         return cls(xyxy=xyxy, mask=mask)
 
     @classmethod
@@ -802,6 +780,69 @@ class Detections:
             class_id=paddledet_result["bbox"][:, 0].astype(int),
         )
 
+    @classmethod
+    def from_lmm(
+        cls, lmm: Union[LMM, str], result: Union[str, dict], **kwargs
+    ) -> Detections:
+        """
+        Creates a Detections object from the given result string based on the specified
+        Large Multimodal Model (LMM).
+
+        Args:
+            lmm (Union[LMM, str]): The type of LMM (Large Multimodal Model) to use.
+            result (str): The result string containing the detection data.
+            **kwargs: Additional keyword arguments required by the specified LMM.
+
+        Returns:
+            Detections: A new Detections object.
+
+        Raises:
+            ValueError: If the LMM is invalid, required arguments are missing, or
+                disallowed arguments are provided.
+            ValueError: If the specified LMM is not supported.
+
+        Examples:
+            ```python
+            import supervision as sv
+
+            paligemma_result = "<loc0256><loc0256><loc0768><loc0768> cat"
+            detections = sv.Detections.from_lmm(
+                sv.LMM.PALIGEMMA,
+                paligemma_result,
+                resolution_wh=(1000, 1000),
+                classes=['cat', 'dog']
+            )
+            detections.xyxy
+            # array([[250., 250., 750., 750.]])
+
+            detections.class_id
+            # array([0])
+            ```
+        """
+        lmm = validate_lmm_parameters(lmm, result, kwargs)
+
+        if lmm == LMM.PALIGEMMA:
+            assert isinstance(result, str)
+            xyxy, class_id, class_name = from_paligemma(result, **kwargs)
+            data = {CLASS_NAME_DATA_FIELD: class_name}
+            return cls(xyxy=xyxy, class_id=class_id, data=data)
+
+        if lmm == LMM.FLORENCE_2:
+            assert isinstance(result, dict)
+            xyxy, labels, mask, xyxyxyxy = from_florence_2(result, **kwargs)
+            if len(xyxy) == 0:
+                return cls.empty()
+
+            data = {}
+            if labels is not None:
+                data[CLASS_NAME_DATA_FIELD] = labels
+            if xyxyxyxy is not None:
+                data[ORIENTED_BOX_COORDINATES] = xyxyxyxy
+
+            return cls(xyxy=xyxy, mask=mask, data=data)
+
+        raise ValueError(f"Unsupported LMM: {lmm}")
+
     @classmethod
     def empty(cls) -> Detections:
         """
@@ -824,6 +865,14 @@ class Detections:
             class_id=np.array([], dtype=int),
         )
 
+    def is_empty(self) -> bool:
+        """
+        Returns `True` if the `Detections` object is considered empty.
+        """
+        empty_detections = Detections.empty()
+        empty_detections.data = self.data
+        return self == empty_detections
+
     @classmethod
     def merge(cls, detections_list: List[Detections]) -> Detections:
         """
@@ -831,9 +880,14 @@ class Detections:
 
         This method takes a list of Detections objects and combines their
         respective fields (`xyxy`, `mask`, `confidence`, `class_id`, and `tracker_id`)
-        into a single Detections object. If all elements in a field are not
-        `None`, the corresponding field will be stacked.
-        Otherwise, the field will be set to `None`.
+        into a single Detections object.
+
+        For example, if merging Detections with 3 and 4 detected objects, this method
+        will return a Detections with 7 objects (7 entries in `xyxy`, `mask`, etc).
+
+        !!! Note
+
+            When merging, empty `Detections` objects are ignored.
 
         Args:
             detections_list (List[Detections]): A list of Detections objects to merge.
@@ -873,6 +927,10 @@ class Detections:
             array([0.1, 0.2, 0.3])
             ```
         """
+        detections_list = [
+            detections for detections in detections_list if not detections.is_empty()
+        ]
+
         if len(detections_list) == 0:
             return Detections.empty()
 
@@ -1100,10 +1158,10 @@ class Detections:
         from a segmentation model, the IoU mask is applied. Otherwise, box IoU is used.
 
         Args:
-            threshold (float, optional): The intersection-over-union threshold
+            threshold (float): The intersection-over-union threshold
                 to use for non-maximum suppression. I'm the lower the value the more
                 restrictive the NMS becomes. Defaults to 0.5.
-            class_agnostic (bool, optional): Whether to perform class-agnostic
+            class_agnostic (bool): Whether to perform class-agnostic
                 non-maximum suppression. If True, the class_id of each detection
                 will be ignored. Defaults to False.
 
@@ -1147,3 +1205,195 @@ class Detections:
             )
 
         return self[indices]
+
+    def with_nmm(
+        self, threshold: float = 0.5, class_agnostic: bool = False
+    ) -> Detections:
+        """
+        Perform non-maximum merging on the current set of object detections.
+
+        Args:
+            threshold (float): The intersection-over-union threshold
+                to use for non-maximum merging. Defaults to 0.5.
+            class_agnostic (bool): Whether to perform class-agnostic
+                non-maximum merging. If True, the class_id of each detection
+                will be ignored. Defaults to False.
+
+        Returns:
+            Detections: A new Detections object containing the subset of detections
+                after non-maximum merging.
+
+        Raises:
+            AssertionError: If `confidence` is None or `class_id` is None and
+                class_agnostic is False.
+
+        ![non-max-merging](https://media.roboflow.com/supervision-docs/non-max-merging.png){ align=center width="800" }
+        """  # noqa: E501 // docs
+        if len(self) == 0:
+            return self
+
+        assert (
+            self.confidence is not None
+        ), "Detections confidence must be given for NMM to be executed."
+
+        if class_agnostic:
+            predictions = np.hstack((self.xyxy, self.confidence.reshape(-1, 1)))
+        else:
+            assert self.class_id is not None, (
+                "Detections class_id must be given for NMM to be executed. If you"
+                " intended to perform class agnostic NMM set class_agnostic=True."
+            )
+            predictions = np.hstack(
+                (
+                    self.xyxy,
+                    self.confidence.reshape(-1, 1),
+                    self.class_id.reshape(-1, 1),
+                )
+            )
+
+        merge_groups = box_non_max_merge(
+            predictions=predictions, iou_threshold=threshold
+        )
+
+        result = []
+        for merge_group in merge_groups:
+            unmerged_detections = [self[i] for i in merge_group]
+            merged_detections = merge_inner_detections_objects(
+                unmerged_detections, threshold
+            )
+            result.append(merged_detections)
+
+        return Detections.merge(result)
+
+
+def merge_inner_detection_object_pair(
+    detections_1: Detections, detections_2: Detections
+) -> Detections:
+    """
+    Merges two Detections object into a single Detections object.
+    Assumes each Detections contains exactly one object.
+
+    A `winning` detection is determined based on the confidence score of the two
+    input detections. This winning detection is then used to specify which
+    `class_id`, `tracker_id`, and `data` to include in the merged Detections object.
+
+    The resulting `confidence` of the merged object is calculated by the weighted
+    contribution of ea detection to the merged object.
+    The bounding boxes and masks of the two input detections are merged into a
+    single bounding box and mask, respectively.
+
+    Args:
+        detections_1 (Detections):
+            The first Detections object
+        detections_2 (Detections):
+            The second Detections object
+
+    Returns:
+        Detections: A new Detections object, with merged attributes.
+
+    Raises:
+        ValueError: If the input Detections objects do not have exactly 1 detected
+            object.
+
+    Example:
+        ```python
+        import cv2
+        import supervision as sv
+        from inference import get_model
+
+        image = cv2.imread(<SOURCE_IMAGE_PATH>)
+        model = get_model(model_id="yolov8s-640")
+
+        result = model.infer(image)[0]
+        detections = sv.Detections.from_inference(result)
+
+        merged_detections = merge_object_detection_pair(
+            detections[0], detections[1])
+        ```
+    """
+    if len(detections_1) != 1 or len(detections_2) != 1:
+        raise ValueError("Both Detections should have exactly 1 detected object.")
+
+    validate_fields_both_defined_or_none(detections_1, detections_2)
+
+    xyxy_1 = detections_1.xyxy[0]
+    xyxy_2 = detections_2.xyxy[0]
+    if detections_1.confidence is None and detections_2.confidence is None:
+        merged_confidence = None
+    else:
+        detection_1_area = (xyxy_1[2] - xyxy_1[0]) * (xyxy_1[3] - xyxy_1[1])
+        detections_2_area = (xyxy_2[2] - xyxy_2[0]) * (xyxy_2[3] - xyxy_2[1])
+        merged_confidence = (
+            detection_1_area * detections_1.confidence[0]
+            + detections_2_area * detections_2.confidence[0]
+        ) / (detection_1_area + detections_2_area)
+        merged_confidence = np.array([merged_confidence])
+
+    merged_x1, merged_y1 = np.minimum(xyxy_1[:2], xyxy_2[:2])
+    merged_x2, merged_y2 = np.maximum(xyxy_1[2:], xyxy_2[2:])
+    merged_xyxy = np.array([[merged_x1, merged_y1, merged_x2, merged_y2]])
+
+    if detections_1.mask is None and detections_2.mask is None:
+        merged_mask = None
+    else:
+        merged_mask = np.logical_or(detections_1.mask, detections_2.mask)
+
+    if detections_1.confidence is None and detections_2.confidence is None:
+        winning_detection = detections_1
+    elif detections_1.confidence[0] >= detections_2.confidence[0]:
+        winning_detection = detections_1
+    else:
+        winning_detection = detections_2
+
+    return Detections(
+        xyxy=merged_xyxy,
+        mask=merged_mask,
+        confidence=merged_confidence,
+        class_id=winning_detection.class_id,
+        tracker_id=winning_detection.tracker_id,
+        data=winning_detection.data,
+    )
+
+
+def merge_inner_detections_objects(
+    detections: List[Detections], threshold=0.5
+) -> Detections:
+    """
+    Given N detections each of length 1 (exactly one object inside), combine them into a
+    single detection object of length 1. The contained inner object will be the merged
+    result of all the input detections.
+
+    For example, this lets you merge N boxes into one big box, N masks into one mask,
+    etc.
+    """
+    detections_1 = detections[0]
+    for detections_2 in detections[1:]:
+        box_iou = box_iou_batch(detections_1.xyxy, detections_2.xyxy)[0]
+        if box_iou < threshold:
+            break
+        detections_1 = merge_inner_detection_object_pair(detections_1, detections_2)
+    return detections_1
+
+
+def validate_fields_both_defined_or_none(
+    detections_1: Detections, detections_2: Detections
+) -> None:
+    """
+    Verify that for each optional field in the Detections, both instances either have
+    the field set to None or both have it set to non-None values.
+
+    `data` field is ignored.
+
+    Raises:
+        ValueError: If one field is None and the other is not, for any of the fields.
+    """
+    attributes = get_instance_variables(detections_1)
+    for attribute in attributes:
+        value_1 = getattr(detections_1, attribute)
+        value_2 = getattr(detections_2, attribute)
+
+        if (value_1 is None) != (value_2 is None):
+            raise ValueError(
+                f"Field '{attribute}' should be consistently None or not None in both "
+                "Detections."
+            )

supervision/detection/line_zone.py

@@ -1,12 +1,15 @@
+import warnings
 from typing import Dict, Iterable, Optional, Tuple
 
 import cv2
 import numpy as np
 
 from supervision.detection.core import Detections
+from supervision.detection.utils import cross_product
 from supervision.draw.color import Color
 from supervision.draw.utils import draw_text
 from supervision.geometry.core import Point, Position, Vector
+from supervision.utils.internal import SupervisionWarnings
 
 
 class LineZone:
@@ -81,6 +84,8 @@ class LineZone:
         self.in_count: int = 0
         self.out_count: int = 0
         self.triggering_anchors = triggering_anchors
+        if not list(self.triggering_anchors):
+            raise ValueError("Triggering anchors cannot be empty.")
 
     @staticmethod
     def calculate_region_of_interest_limits(vector: Vector) -> Tuple[Vector, Vector]:
@@ -140,6 +145,15 @@ class LineZone:
         if len(detections) == 0:
             return crossed_in, crossed_out
 
+        if detections.tracker_id is None:
+            warnings.warn(
+                "Line zone counting skipped. LineZone requires tracker_id. Refer to "
+                "https://supervision.roboflow.com/latest/trackers for more "
+                "information.",
+                category=SupervisionWarnings,
+            )
+            return crossed_in, crossed_out
+
         all_anchors = np.array(
             [
                 detections.get_anchors_coordinates(anchor)
@@ -147,31 +161,23 @@ class LineZone:
             ]
         )
 
+        cross_products_1 = cross_product(all_anchors, self.limits[0])
+        cross_products_2 = cross_product(all_anchors, self.limits[1])
+        in_limits = (cross_products_1 > 0) == (cross_products_2 > 0)
+        in_limits = np.all(in_limits, axis=0)
+
+        triggers = cross_product(all_anchors, self.vector) < 0
+        has_any_left_trigger = np.any(triggers, axis=0)
+        has_any_right_trigger = np.any(~triggers, axis=0)
+        is_uniformly_triggered = ~(has_any_left_trigger & has_any_right_trigger)
         for i, tracker_id in enumerate(detections.tracker_id):
-            if tracker_id is None:
+            if not in_limits[i]:
                 continue
 
-            box_anchors = [Point(x=x, y=y) for x, y in all_anchors[:, i, :]]
-
-            in_limits = all(
-                [
-                    self.is_point_in_limits(point=anchor, limits=self.limits)
-                    for anchor in box_anchors
-                ]
-            )
-
-            if not in_limits:
+            if not is_uniformly_triggered[i]:
                 continue
 
-            triggers = [
-                self.vector.cross_product(point=anchor) < 0 for anchor in box_anchors
-            ]
-
-            if len(set(triggers)) == 2:
-                continue
-
-            tracker_state = triggers[0]
-
+            tracker_state = has_any_left_trigger[i]
             if tracker_id not in self.tracker_state:
                 self.tracker_state[tracker_id] = tracker_state
                 continue

supervision/detection/lmm.py

@@ -0,0 +1,184 @@
+import re
+from enum import Enum
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import numpy as np
+
+from supervision.detection.utils import polygon_to_mask, polygon_to_xyxy
+
+
+class LMM(Enum):
+    PALIGEMMA = "paligemma"
+    FLORENCE_2 = "florence_2"
+
+
+RESULT_TYPES: Dict[LMM, type] = {LMM.PALIGEMMA: str, LMM.FLORENCE_2: dict}
+
+REQUIRED_ARGUMENTS: Dict[LMM, List[str]] = {
+    LMM.PALIGEMMA: ["resolution_wh"],
+    LMM.FLORENCE_2: ["resolution_wh"],
+}
+
+ALLOWED_ARGUMENTS: Dict[LMM, List[str]] = {
+    LMM.PALIGEMMA: ["resolution_wh", "classes"],
+    LMM.FLORENCE_2: ["resolution_wh"],
+}
+
+SUPPORTED_TASKS_FLORENCE_2 = [
+    "<OD>",
+    "<CAPTION_TO_PHRASE_GROUNDING>",
+    "<DENSE_REGION_CAPTION>",
+    "<REGION_PROPOSAL>",
+    "<OCR_WITH_REGION>",
+    "<REFERRING_EXPRESSION_SEGMENTATION>",
+    "<REGION_TO_SEGMENTATION>",
+    "<OPEN_VOCABULARY_DETECTION>",
+    "<REGION_TO_CATEGORY>",
+    "<REGION_TO_DESCRIPTION>",
+]
+
+
+def validate_lmm_parameters(
+    lmm: Union[LMM, str], result: Any, kwargs: Dict[str, Any]
+) -> LMM:
+    if isinstance(lmm, str):
+        try:
+            lmm = LMM(lmm.lower())
+        except ValueError:
+            raise ValueError(
+                f"Invalid lmm value: {lmm}. Must be one of {[e.value for e in LMM]}"
+            )
+
+    if not isinstance(result, RESULT_TYPES[lmm]):
+        raise ValueError(
+            f"Invalid LMM result type: {type(result)}. Must be {RESULT_TYPES[lmm]}"
+        )
+
+    required_args = REQUIRED_ARGUMENTS.get(lmm, [])
+    for arg in required_args:
+        if arg not in kwargs:
+            raise ValueError(f"Missing required argument: {arg}")
+
+    allowed_args = ALLOWED_ARGUMENTS.get(lmm, [])
+    for arg in kwargs:
+        if arg not in allowed_args:
+            raise ValueError(f"Argument {arg} is not allowed for {lmm.name}")
+
+    return lmm
+
+
+def from_paligemma(
+    result: str, resolution_wh: Tuple[int, int], classes: Optional[List[str]] = None
+) -> Tuple[np.ndarray, Optional[np.ndarray], np.ndarray]:
+    w, h = resolution_wh
+    pattern = re.compile(
+        r"(?<!<loc\d{4}>)<loc(\d{4})><loc(\d{4})><loc(\d{4})><loc(\d{4})> ([\w\s\-]+)"
+    )
+    matches = pattern.findall(result)
+    matches = np.array(matches) if matches else np.empty((0, 5))
+
+    xyxy, class_name = matches[:, [1, 0, 3, 2]], matches[:, 4]
+    xyxy = xyxy.astype(int) / 1024 * np.array([w, h, w, h])
+    class_name = np.char.strip(class_name.astype(str))
+    class_id = None
+
+    if classes is not None:
+        mask = np.array([name in classes for name in class_name]).astype(bool)
+        xyxy, class_name = xyxy[mask], class_name[mask]
+        class_id = np.array([classes.index(name) for name in class_name])
+
+    return xyxy, class_id, class_name
+
+
+def from_florence_2(
+    result: dict, resolution_wh: Tuple[int, int]
+) -> Tuple[
+    np.ndarray, Optional[np.ndarray], Optional[np.ndarray], Optional[np.ndarray]
+]:
+    """
+    Parse results from the Florence 2 multi-model model.
+    https://huggingface.co/microsoft/Florence-2-large
+
+    Parameters:
+        result: dict containing the model output
+
+    Returns:
+        xyxy (np.ndarray): An array of shape `(n, 4)` containing
+            the bounding boxes coordinates in format `[x1, y1, x2, y2]`
+        labels: (Optional[np.ndarray]): An array of shape `(n,)` containing
+            the class labels for each bounding box
+        masks: (Optional[np.ndarray]): An array of shape `(n, h, w)` containing
+            the segmentation masks for each bounding box
+        obb_boxes: (Optional[np.ndarray]): An array of shape `(n, 4, 2)` containing
+            oriented bounding boxes.
+    """
+    assert len(result) == 1, f"Expected result with a single element. Got: {result}"
+    task = list(result.keys())[0]
+    if task not in SUPPORTED_TASKS_FLORENCE_2:
+        raise ValueError(
+            f"{task} not supported. Supported tasks are: {SUPPORTED_TASKS_FLORENCE_2}"
+        )
+    result = result[task]
+
+    if task in ["<OD>", "<CAPTION_TO_PHRASE_GROUNDING>", "<DENSE_REGION_CAPTION>"]:
+        xyxy = np.array(result["bboxes"], dtype=np.float32)
+        labels = np.array(result["labels"])
+        return xyxy, labels, None, None
+
+    if task == "<REGION_PROPOSAL>":
+        xyxy = np.array(result["bboxes"], dtype=np.float32)
+        # provides labels, but they are ["", "", "", ...]
+        return xyxy, None, None, None
+
+    if task == "<OCR_WITH_REGION>":
+        xyxyxyxy = np.array(result["quad_boxes"], dtype=np.float32)
+        xyxyxyxy = xyxyxyxy.reshape(-1, 4, 2)
+        xyxy = np.array([polygon_to_xyxy(polygon) for polygon in xyxyxyxy])
+        labels = np.array(result["labels"])
+        return xyxy, labels, None, xyxyxyxy
+
+    if task in ["<REFERRING_EXPRESSION_SEGMENTATION>", "<REGION_TO_SEGMENTATION>"]:
+        xyxy_list = []
+        masks_list = []
+        for polygons_of_same_class in result["polygons"]:
+            for polygon in polygons_of_same_class:
+                polygon = np.reshape(polygon, (-1, 2)).astype(np.int32)
+                mask = polygon_to_mask(polygon, resolution_wh).astype(bool)
+                masks_list.append(mask)
+                xyxy = polygon_to_xyxy(polygon)
+                xyxy_list.append(xyxy)
+            # per-class labels also provided, but they are ["", "", "", ...]
+            # when we figure out how to set class names, we can do
+            # zip(result["labels"], result["polygons"])
+        xyxy = np.array(xyxy_list, dtype=np.float32)
+        masks = np.array(masks_list)
+        return xyxy, None, masks, None
+
+    if task == "<OPEN_VOCABULARY_DETECTION>":
+        xyxy = np.array(result["bboxes"], dtype=np.float32)
+        labels = np.array(result["bboxes_labels"])
+        # Also has "polygons" and "polygons_labels", but they don't seem to be used
+        return xyxy, labels, None, None
+
+    if task in ["<REGION_TO_CATEGORY>", "<REGION_TO_DESCRIPTION>"]:
+        assert isinstance(
+            result, str
+        ), f"Expected string as <REGION_TO_CATEGORY> result, got {type(result)}"
+
+        if result == "No object detected.":
+            return np.empty((0, 4), dtype=np.float32), np.array([]), None, None
+
+        pattern = re.compile(r"<loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)>")
+        match = pattern.search(result)
+        assert (
+            match is not None
+        ), f"Expected string to end in location tags, but got {result}"
+
+        w, h = resolution_wh
+        xyxy = np.array([match.groups()], dtype=np.float32)
+        xyxy *= np.array([w, h, w, h]) / 1000
+        result_string = result[: match.start()]
+        labels = np.array([result_string])
+        return xyxy, labels, None, None
+
+    assert False, f"Unimplemented task: {task}"

supervision/detection/overlap_filter.py

@@ -0,0 +1,270 @@
+from __future__ import annotations
+
+from enum import Enum
+from typing import List, Union
+
+import numpy as np
+import numpy.typing as npt
+
+from supervision.detection.utils import box_iou_batch, mask_iou_batch
+
+
+def resize_masks(masks: np.ndarray, max_dimension: int = 640) -> np.ndarray:
+    """
+    Resize all masks in the array to have a maximum dimension of max_dimension,
+    maintaining aspect ratio.
+
+    Args:
+        masks (np.ndarray): 3D array of binary masks with shape (N, H, W).
+        max_dimension (int): The maximum dimension for the resized masks.
+
+    Returns:
+        np.ndarray: Array of resized masks.
+    """
+    max_height = np.max(masks.shape[1])
+    max_width = np.max(masks.shape[2])
+    scale = min(max_dimension / max_height, max_dimension / max_width)
+
+    new_height = int(scale * max_height)
+    new_width = int(scale * max_width)
+
+    x = np.linspace(0, max_width - 1, new_width).astype(int)
+    y = np.linspace(0, max_height - 1, new_height).astype(int)
+    xv, yv = np.meshgrid(x, y)
+
+    resized_masks = masks[:, yv, xv]
+
+    resized_masks = resized_masks.reshape(masks.shape[0], new_height, new_width)
+    return resized_masks
+
+
+def mask_non_max_suppression(
+    predictions: np.ndarray,
+    masks: np.ndarray,
+    iou_threshold: float = 0.5,
+    mask_dimension: int = 640,
+) -> np.ndarray:
+    """
+    Perform Non-Maximum Suppression (NMS) on segmentation predictions.
+
+    Args:
+        predictions (np.ndarray): A 2D array of object detection predictions in
+            the format of `(x_min, y_min, x_max, y_max, score)`
+            or `(x_min, y_min, x_max, y_max, score, class)`. Shape: `(N, 5)` or
+            `(N, 6)`, where N is the number of predictions.
+        masks (np.ndarray): A 3D array of binary masks corresponding to the predictions.
+            Shape: `(N, H, W)`, where N is the number of predictions, and H, W are the
+            dimensions of each mask.
+        iou_threshold (float): The intersection-over-union threshold
+            to use for non-maximum suppression.
+        mask_dimension (int): The dimension to which the masks should be
+            resized before computing IOU values. Defaults to 640.
+
+    Returns:
+        np.ndarray: A boolean array indicating which predictions to keep after
+            non-maximum suppression.
+
+    Raises:
+        AssertionError: If `iou_threshold` is not within the closed
+        range from `0` to `1`.
+    """
+    assert 0 <= iou_threshold <= 1, (
+        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
+        f"{iou_threshold} given."
+    )
+    rows, columns = predictions.shape
+
+    if columns == 5:
+        predictions = np.c_[predictions, np.zeros(rows)]
+
+    sort_index = predictions[:, 4].argsort()[::-1]
+    predictions = predictions[sort_index]
+    masks = masks[sort_index]
+    masks_resized = resize_masks(masks, mask_dimension)
+    ious = mask_iou_batch(masks_resized, masks_resized)
+    categories = predictions[:, 5]
+
+    keep = np.ones(rows, dtype=bool)
+    for i in range(rows):
+        if keep[i]:
+            condition = (ious[i] > iou_threshold) & (categories[i] == categories)
+            keep[i + 1 :] = np.where(condition[i + 1 :], False, keep[i + 1 :])
+
+    return keep[sort_index.argsort()]
+
+
+def box_non_max_suppression(
+    predictions: np.ndarray, iou_threshold: float = 0.5
+) -> np.ndarray:
+    """
+    Perform Non-Maximum Suppression (NMS) on object detection predictions.
+
+    Args:
+        predictions (np.ndarray): An array of object detection predictions in
+            the format of `(x_min, y_min, x_max, y_max, score)`
+            or `(x_min, y_min, x_max, y_max, score, class)`.
+        iou_threshold (float): The intersection-over-union threshold
+            to use for non-maximum suppression.
+
+    Returns:
+        np.ndarray: A boolean array indicating which predictions to keep after n
+            on-maximum suppression.
+
+    Raises:
+        AssertionError: If `iou_threshold` is not within the
+            closed range from `0` to `1`.
+    """
+    assert 0 <= iou_threshold <= 1, (
+        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
+        f"{iou_threshold} given."
+    )
+    rows, columns = predictions.shape
+
+    # add column #5 - category filled with zeros for agnostic nms
+    if columns == 5:
+        predictions = np.c_[predictions, np.zeros(rows)]
+
+    # sort predictions column #4 - score
+    sort_index = np.flip(predictions[:, 4].argsort())
+    predictions = predictions[sort_index]
+
+    boxes = predictions[:, :4]
+    categories = predictions[:, 5]
+    ious = box_iou_batch(boxes, boxes)
+    ious = ious - np.eye(rows)
+
+    keep = np.ones(rows, dtype=bool)
+
+    for index, (iou, category) in enumerate(zip(ious, categories)):
+        if not keep[index]:
+            continue
+
+        # drop detections with iou > iou_threshold and
+        # same category as current detections
+        condition = (iou > iou_threshold) & (categories == category)
+        keep = keep & ~condition
+
+    return keep[sort_index.argsort()]
+
+
+def group_overlapping_boxes(
+    predictions: npt.NDArray[np.float64], iou_threshold: float = 0.5
+) -> List[List[int]]:
+    """
+    Apply greedy version of non-maximum merging to avoid detecting too many
+    overlapping bounding boxes for a given object.
+
+    Args:
+        predictions (npt.NDArray[np.float64]): An array of shape `(n, 5)` containing
+            the bounding boxes coordinates in format `[x1, y1, x2, y2]`
+            and the confidence scores.
+        iou_threshold (float): The intersection-over-union threshold
+            to use for non-maximum suppression. Defaults to 0.5.
+
+    Returns:
+        List[List[int]]: Groups of prediction indices be merged.
+            Each group may have 1 or more elements.
+    """
+    merge_groups: List[List[int]] = []
+
+    scores = predictions[:, 4]
+    order = scores.argsort()
+
+    while len(order) > 0:
+        idx = int(order[-1])
+
+        order = order[:-1]
+        if len(order) == 0:
+            merge_groups.append([idx])
+            break
+
+        merge_candidate = np.expand_dims(predictions[idx], axis=0)
+        ious = box_iou_batch(predictions[order][:, :4], merge_candidate[:, :4])
+        ious = ious.flatten()
+
+        above_threshold = ious >= iou_threshold
+        merge_group = [idx] + np.flip(order[above_threshold]).tolist()
+        merge_groups.append(merge_group)
+        order = order[~above_threshold]
+    return merge_groups
+
+
+def box_non_max_merge(
+    predictions: npt.NDArray[np.float64],
+    iou_threshold: float = 0.5,
+) -> List[List[int]]:
+    """
+    Apply greedy version of non-maximum merging per category to avoid detecting
+    too many overlapping bounding boxes for a given object.
+
+    Args:
+        predictions (npt.NDArray[np.float64]): An array of shape `(n, 5)` or `(n, 6)`
+            containing the bounding boxes coordinates in format `[x1, y1, x2, y2]`,
+            the confidence scores and class_ids. Omit class_id column to allow
+            detections of different classes to be merged.
+        iou_threshold (float): The intersection-over-union threshold
+            to use for non-maximum suppression. Defaults to 0.5.
+
+    Returns:
+        List[List[int]]: Groups of prediction indices be merged.
+            Each group may have 1 or more elements.
+    """
+    if predictions.shape[1] == 5:
+        return group_overlapping_boxes(predictions, iou_threshold)
+
+    category_ids = predictions[:, 5]
+    merge_groups = []
+    for category_id in np.unique(category_ids):
+        curr_indices = np.where(category_ids == category_id)[0]
+        merge_class_groups = group_overlapping_boxes(
+            predictions[curr_indices], iou_threshold
+        )
+
+        for merge_class_group in merge_class_groups:
+            merge_groups.append(curr_indices[merge_class_group].tolist())
+
+    for merge_group in merge_groups:
+        if len(merge_group) == 0:
+            raise ValueError(
+                f"Empty group detected when non-max-merging "
+                f"detections: {merge_groups}"
+            )
+    return merge_groups
+
+
+class OverlapFilter(Enum):
+    """
+    Enum specifying the strategy for filtering overlapping detections.
+
+    Attributes:
+        NONE: Do not filter detections based on overlap.
+        NON_MAX_SUPPRESSION: Filter detections using non-max suppression. This means,
+            detections that overlap by more than a set threshold will be discarded,
+            except for the one with the highest confidence.
+        NON_MAX_MERGE: Merge detections with non-max merging. This means,
+            detections that overlap by more than a set threshold will be merged
+            into a single detection.
+    """
+
+    NONE = "none"
+    NON_MAX_SUPPRESSION = "non_max_suppression"
+    NON_MAX_MERGE = "non_max_merge"
+
+    @classmethod
+    def list(cls):
+        return list(map(lambda c: c.value, cls))
+
+    @classmethod
+    def from_value(cls, value: Union[OverlapFilter, str]) -> OverlapFilter:
+        if isinstance(value, cls):
+            return value
+        if isinstance(value, str):
+            value = value.lower()
+            try:
+                return cls(value)
+            except ValueError:
+                raise ValueError(f"Invalid value: {value}. Must be one of {cls.list()}")
+        raise ValueError(
+            f"Invalid value type: {type(value)}. Must be an instance of "
+            f"{cls.__name__} or str."
+        )

supervision/detection/tools/inference_slicer.py

@@ -1,23 +1,51 @@
+import warnings
 from concurrent.futures import ThreadPoolExecutor, as_completed
-from typing import Callable, Optional, Tuple
+from typing import Callable, Optional, Tuple, Union
 
 import numpy as np
 
+from supervision.config import ORIENTED_BOX_COORDINATES
 from supervision.detection.core import Detections
-from supervision.detection.utils import move_boxes
+from supervision.detection.overlap_filter import OverlapFilter
+from supervision.detection.utils import move_boxes, move_masks, move_oriented_boxes
 from supervision.utils.image import crop_image
+from supervision.utils.internal import (
+    SupervisionWarnings,
+    deprecated_parameter,
+    warn_deprecated,
+)
 
 
-def move_detections(detections: Detections, offset: np.array) -> Detections:
+def move_detections(
+    detections: Detections,
+    offset: np.ndarray,
+    resolution_wh: Optional[Tuple[int, int]] = None,
+) -> Detections:
     """
     Args:
         detections (sv.Detections): Detections object to be moved.
-        offset (np.array): An array of shape `(2,)` containing offset values in format
+        offset (np.ndarray): An array of shape `(2,)` containing offset values in format
             is `[dx, dy]`.
+        resolution_wh (Tuple[int, int]): The width and height of the desired mask
+            resolution. Required for segmentation detections.
+
     Returns:
         (sv.Detections) repositioned Detections object.
     """
     detections.xyxy = move_boxes(xyxy=detections.xyxy, offset=offset)
+    if ORIENTED_BOX_COORDINATES in detections.data:
+        detections.data[ORIENTED_BOX_COORDINATES] = move_oriented_boxes(
+            xyxyxyxy=detections.data[ORIENTED_BOX_COORDINATES], offset=offset
+        )
+    if detections.mask is not None:
+        if resolution_wh is None:
+            raise ValueError(
+                "Resolution width and height are required for moving segmentation "
+                "detections. This should be the same as (width, height) of image shape."
+            )
+        detections.mask = move_masks(
+            masks=detections.mask, offset=offset, resolution_wh=resolution_wh
+        )
     return detections
 
 
@@ -30,12 +58,19 @@ class InferenceSlicer:
     slice, and then merging the detections.
 
     Args:
-        slice_wh (Tuple[int, int]): Dimensions of each slice in the format
-            `(width, height)`.
-        overlap_ratio_wh (Tuple[float, float]): Overlap ratio between consecutive
-            slices in the format `(width_ratio, height_ratio)`.
-        iou_threshold (Optional[float]): Intersection over Union (IoU) threshold
-            used for non-max suppression.
+        slice_wh (Tuple[int, int]): Dimensions of each slice measured in pixels. The
+            tuple should be in the format `(width, height)`.
+        overlap_ratio_wh (Optional[Tuple[float, float]]): A tuple representing the
+            desired overlap ratio for width and height between consecutive slices.
+            Each value should be in the range [0, 1), where 0 means no overlap and
+            a value close to 1 means high overlap.
+        overlap_wh (Optional[Tuple[int, int]]): A tuple representing the desired
+            overlap for width and height between consecutive slices measured in pixels.
+            Each value should be greater than or equal to 0.
+        overlap_filter (Union[OverlapFilter, str]): Strategy for
+            filtering or merging overlapping detections in slices.
+        iou_threshold (float): Intersection over Union (IoU) threshold
+            used when filtering by overlap.
         callback (Callable): A function that performs inference on a given image
             slice and returns detections.
         thread_workers (int): Number of threads for parallel execution.
@@ -47,17 +82,37 @@ class InferenceSlicer:
         not a multiple of the slice's width or height minus the overlap.
     """
 
+    @deprecated_parameter(
+        old_parameter="overlap_filter_strategy",
+        new_parameter="overlap_filter",
+        map_function=lambda x: x,
+        warning_message="`{old_parameter}` in `{function_name}` is deprecated and will "
+        "be removed in `supervision-0.27.0`. Use '{new_parameter}' "
+        "instead.",
+    )
     def __init__(
         self,
         callback: Callable[[np.ndarray], Detections],
         slice_wh: Tuple[int, int] = (320, 320),
-        overlap_ratio_wh: Tuple[float, float] = (0.2, 0.2),
-        iou_threshold: Optional[float] = 0.5,
+        overlap_ratio_wh: Optional[Tuple[float, float]] = (0.2, 0.2),
+        overlap_wh: Optional[Tuple[int, int]] = None,
+        overlap_filter: Union[OverlapFilter, str] = OverlapFilter.NON_MAX_SUPPRESSION,
+        iou_threshold: float = 0.5,
         thread_workers: int = 1,
     ):
-        self.slice_wh = slice_wh
+        if overlap_ratio_wh is not None:
+            warn_deprecated(
+                "`overlap_ratio_wh` in `InferenceSlicer.__init__` is deprecated and "
+                "will be removed in `supervision-0.27.0`. Use `overlap_wh` instead."
+            )
+
+        self._validate_overlap(overlap_ratio_wh, overlap_wh)
         self.overlap_ratio_wh = overlap_ratio_wh
+        self.overlap_wh = overlap_wh
+
+        self.slice_wh = slice_wh
         self.iou_threshold = iou_threshold
+        self.overlap_filter = OverlapFilter.from_value(overlap_filter)
         self.callback = callback
         self.thread_workers = thread_workers
 
@@ -88,7 +143,10 @@ class InferenceSlicer:
                 result = model(image_slice)[0]
                 return sv.Detections.from_ultralytics(result)
 
-            slicer = sv.InferenceSlicer(callback = callback)
+            slicer = sv.InferenceSlicer(
+                callback=callback,
+                overlap_filter_strategy=sv.OverlapFilter.NON_MAX_SUPPRESSION,
+            )
 
             detections = slicer(image)
             ```
@@ -99,6 +157,7 @@ class InferenceSlicer:
             resolution_wh=resolution_wh,
             slice_wh=self.slice_wh,
             overlap_ratio_wh=self.overlap_ratio_wh,
+            overlap_wh=self.overlap_wh,
         )
 
         with ThreadPoolExecutor(max_workers=self.thread_workers) as executor:
@@ -108,9 +167,19 @@ class InferenceSlicer:
             for future in as_completed(futures):
                 detections_list.append(future.result())
 
-        return Detections.merge(detections_list=detections_list).with_nms(
-            threshold=self.iou_threshold
-        )
+        merged = Detections.merge(detections_list=detections_list)
+        if self.overlap_filter == OverlapFilter.NONE:
+            return merged
+        elif self.overlap_filter == OverlapFilter.NON_MAX_SUPPRESSION:
+            return merged.with_nms(threshold=self.iou_threshold)
+        elif self.overlap_filter == OverlapFilter.NON_MAX_MERGE:
+            return merged.with_nmm(threshold=self.iou_threshold)
+        else:
+            warnings.warn(
+                f"Invalid overlap filter strategy: {self.overlap_filter}",
+                category=SupervisionWarnings,
+            )
+            return merged
 
     def _run_callback(self, image, offset) -> Detections:
         """
@@ -126,7 +195,10 @@ class InferenceSlicer:
         """
         image_slice = crop_image(image=image, xyxy=offset)
         detections = self.callback(image_slice)
-        detections = move_detections(detections=detections, offset=offset[:2])
+        resolution_wh = (image.shape[1], image.shape[0])
+        detections = move_detections(
+            detections=detections, offset=offset[:2], resolution_wh=resolution_wh
+        )
 
         return detections
 
@@ -134,7 +206,8 @@ class InferenceSlicer:
     def _generate_offset(
         resolution_wh: Tuple[int, int],
         slice_wh: Tuple[int, int],
-        overlap_ratio_wh: Tuple[float, float],
+        overlap_ratio_wh: Optional[Tuple[float, float]],
+        overlap_wh: Optional[Tuple[int, int]],
     ) -> np.ndarray:
         """
         Generate offset coordinates for slicing an image based on the given resolution,
@@ -143,12 +216,15 @@ class InferenceSlicer:
         Args:
             resolution_wh (Tuple[int, int]): A tuple representing the width and height
                 of the image to be sliced.
-            slice_wh (Tuple[int, int]): A tuple representing the desired width and
-                height of each slice.
-            overlap_ratio_wh (Tuple[float, float]): A tuple representing the desired
-                overlap ratio for width and height between consecutive slices. Each
-                value should be in the range [0, 1), where 0 means no overlap and a
-                value close to 1 means high overlap.
+            slice_wh (Tuple[int, int]): Dimensions of each slice measured in pixels. The
+            tuple should be in the format `(width, height)`.
+            overlap_ratio_wh (Optional[Tuple[float, float]]): A tuple representing the
+                desired overlap ratio for width and height between consecutive slices.
+                Each value should be in the range [0, 1), where 0 means no overlap and
+                a value close to 1 means high overlap.
+            overlap_wh (Optional[Tuple[int, int]]): A tuple representing the desired
+                overlap for width and height between consecutive slices measured in
+                pixels. Each value should be greater than or equal to 0.
 
         Returns:
             np.ndarray: An array of shape `(n, 4)` containing coordinates for each
@@ -163,10 +239,19 @@ class InferenceSlicer:
         """
         slice_width, slice_height = slice_wh
         image_width, image_height = resolution_wh
-        overlap_ratio_width, overlap_ratio_height = overlap_ratio_wh
+        overlap_width = (
+            overlap_wh[0]
+            if overlap_wh is not None
+            else int(overlap_ratio_wh[0] * slice_width)
+        )
+        overlap_height = (
+            overlap_wh[1]
+            if overlap_wh is not None
+            else int(overlap_ratio_wh[1] * slice_height)
+        )
 
-        width_stride = slice_width - int(overlap_ratio_width * slice_width)
-        height_stride = slice_height - int(overlap_ratio_height * slice_height)
+        width_stride = slice_width - overlap_width
+        height_stride = slice_height - overlap_height
 
         ws = np.arange(0, image_width, width_stride)
         hs = np.arange(0, image_height, height_stride)
@@ -178,3 +263,32 @@ class InferenceSlicer:
         offsets = np.stack([xmin, ymin, xmax, ymax], axis=-1).reshape(-1, 4)
 
         return offsets
+
+    @staticmethod
+    def _validate_overlap(
+        overlap_ratio_wh: Optional[Tuple[float, float]],
+        overlap_wh: Optional[Tuple[int, int]],
+    ) -> None:
+        if overlap_ratio_wh is not None and overlap_wh is not None:
+            raise ValueError(
+                "Both `overlap_ratio_wh` and `overlap_wh` cannot be provided. "
+                "Please provide only one of them."
+            )
+        if overlap_ratio_wh is None and overlap_wh is None:
+            raise ValueError(
+                "Either `overlap_ratio_wh` or `overlap_wh` must be provided. "
+                "Please provide one of them."
+            )
+
+        if overlap_ratio_wh is not None:
+            if not (0 <= overlap_ratio_wh[0] < 1 and 0 <= overlap_ratio_wh[1] < 1):
+                raise ValueError(
+                    "Overlap ratios must be in the range [0, 1). "
+                    f"Received: {overlap_ratio_wh}"
+                )
+        if overlap_wh is not None:
+            if not (overlap_wh[0] >= 0 and overlap_wh[1] >= 0):
+                raise ValueError(
+                    "Overlap values must be greater than or equal to 0. "
+                    f"Received: {overlap_wh}"
+                )

supervision/detection/tools/polygon_zone.py

@@ -12,7 +12,7 @@ from supervision.draw.color import Color
 from supervision.draw.utils import draw_polygon, draw_text
 from supervision.geometry.core import Position
 from supervision.geometry.utils import get_polygon_center
-from supervision.utils.internal import SupervisionWarnings, deprecated_parameter
+from supervision.utils.internal import SupervisionWarnings
 
 
 class PolygonZone:
@@ -30,14 +30,6 @@ class PolygonZone:
         mask (np.ndarray): The 2D bool mask for the polygon zone
     """
 
-    @deprecated_parameter(
-        old_parameter="triggering_position",
-        new_parameter="triggering_anchors",
-        map_function=lambda x: [x],
-        warning_message="`{old_parameter}` in `{function_name}` is deprecated and will "
-        "be remove in `supervision-0.23.0`. Use '{new_parameter}' "
-        "instead.",
-    )
     def __init__(
         self,
         polygon: npt.NDArray[np.int64],
@@ -54,6 +46,8 @@ class PolygonZone:
 
         self.polygon = polygon.astype(int)
         self.triggering_anchors = triggering_anchors
+        if not list(self.triggering_anchors):
+            raise ValueError("Triggering anchors cannot be empty.")
 
         self.current_count = 0
 
@@ -145,7 +139,7 @@ class PolygonZoneAnnotator:
 
         Parameters:
             scene (np.ndarray): The image on which the polygon zone will be annotated
-            label (Optional[str]): An optional label for the count of detected objects
+            label (Optional[str]): A label for the count of detected objects
                 within the polygon zone (default: None)
 
         Returns:

supervision/detection/tools/smoother.py

@@ -1,3 +1,4 @@
+import warnings
 from collections import defaultdict, deque
 from copy import deepcopy
 from typing import Optional
@@ -5,6 +6,7 @@ from typing import Optional
 import numpy as np
 
 from supervision.detection.core import Detections
+from supervision.utils.internal import SupervisionWarnings
 
 
 class DetectionsSmoother:
@@ -39,7 +41,7 @@ class DetectionsSmoother:
         tracker = sv.ByteTrack(frame_rate=video_info.fps)
         smoother = sv.DetectionsSmoother()
 
-        annotator = sv.BoundingBoxAnnotator()
+        box_annotator = sv.BoxAnnotator()
 
         with sv.VideoSink(<TARGET_FILE_PATH>, video_info=video_info) as sink:
             for frame in frame_generator:
@@ -48,7 +50,7 @@ class DetectionsSmoother:
                 detections = tracker.update_with_detections(detections)
                 detections = smoother.update_with_detections(detections)
 
-                annotated_frame = bounding_box_annotator.annotate(frame.copy(), detections)
+                annotated_frame = box_annotator.annotate(frame.copy(), detections)
                 sink.write_frame(annotated_frame)
         ```
     """  # noqa: E501 // docs
@@ -70,16 +72,16 @@ class DetectionsSmoother:
         """
 
         if detections.tracker_id is None:
-            print(
+            warnings.warn(
                 "Smoothing skipped. DetectionsSmoother requires tracker_id. Refer to "
-                "https://supervision.roboflow.com/latest/trackers for more information."
+                "https://supervision.roboflow.com/latest/trackers for more "
+                "information.",
+                category=SupervisionWarnings,
             )
             return detections
 
         for detection_idx in range(len(detections)):
             tracker_id = detections.tracker_id[detection_idx]
-            if tracker_id is None:
-                continue
 
             self.tracks[tracker_id].append(detections[detection_idx])
 

supervision/detection/tools/transformers.py

@@ -0,0 +1,247 @@
+import io
+from typing import Any, Dict, Optional
+
+import numpy as np
+from PIL import Image
+
+from supervision.config import CLASS_NAME_DATA_FIELD
+from supervision.detection.utils import mask_to_xyxy
+
+
+def process_transformers_detection_result(
+    detection_result: dict, id2label: Optional[Dict[int, str]]
+) -> dict:
+    """
+    Process the result of Transformers object detection functions such as
+    `post_process` (v4) and `post_process_detection` (v5).
+
+    Args:
+        detection_result (dict): Dictionary containing detection results with keys
+            'boxes', 'labels', and 'scores'.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+
+    Returns:
+        dict: Processed detection result including bounding boxes, confidence scores,
+              class IDs, and data.
+    """
+    class_ids = detection_result["labels"].cpu().detach().numpy().astype(int)
+    data = append_class_names_to_data(class_ids, id2label, {})
+
+    return dict(
+        xyxy=detection_result["boxes"].cpu().detach().numpy(),
+        confidence=detection_result["scores"].cpu().detach().numpy(),
+        class_id=class_ids,
+        data=data,
+    )
+
+
+def process_transformers_v4_segmentation_result(
+    segmentation_result: dict, id2label: Optional[Dict[int, str]]
+) -> dict:
+    """
+    Process the result of Transformers segmentation functions such as
+    `post_process_panoptic`, `post_process_segmentation`, and `post_process_instance`
+    (v4).
+
+    Args:
+        segmentation_result (dict): Dictionary containing segmentation results with keys
+            'masks', 'labels', and 'scores'.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+
+    Returns:
+        dict: Processed segmentation result including bounding boxes, masks, confidence
+              scores, class IDs, and data.
+    """
+    if "png_string" in segmentation_result:
+        return process_transformers_v4_panoptic_segmentation_result(
+            segmentation_result, id2label
+        )
+    else:
+        boxes = None
+        if "boxes" in segmentation_result:
+            boxes = segmentation_result["boxes"].cpu().detach().numpy()
+        masks = segmentation_result["masks"].cpu().detach().numpy().astype(bool)
+        class_ids = segmentation_result["labels"].cpu().detach().numpy().astype(int)
+
+        return dict(
+            xyxy=boxes if boxes is not None else mask_to_xyxy(masks),
+            mask=np.squeeze(masks, axis=1) if boxes is not None else masks,
+            confidence=segmentation_result["scores"].cpu().detach().numpy(),
+            class_id=class_ids,
+            data=append_class_names_to_data(class_ids, id2label, {}),
+        )
+
+
+def process_transformers_v5_segmentation_result(
+    segmentation_result: dict, id2label: Optional[Dict[int, str]]
+) -> dict:
+    """
+    Process the result of Transformers segmentation functions such as
+    `post_process_semantic_segmentation`, `post_process_instance_segmentation`, and
+    `post_process_panoptic_segmentation` (v5).
+
+    Args:
+        segmentation_result (Union[dict, np.ndarray]): Either a dictionary containing
+            segmentation results or an ndarray representing a segmentation map.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+
+    Returns:
+        dict: Processed segmentation result including bounding boxes, masks, confidence
+              scores, class IDs, and data.
+    """
+    if segmentation_result.__class__.__name__ == "Tensor":
+        segmentation_array = segmentation_result.cpu().detach().numpy()
+        return process_transformers_v5_panoptic_segmentation_result(
+            segmentation_array, id2label
+        )
+
+    return process_transformers_v5_semantic_or_instance_segmentation_result(
+        segmentation_result, id2label
+    )
+
+
+def process_transformers_v5_semantic_or_instance_segmentation_result(
+    segmentation_result: dict, id2label: Optional[Dict[int, str]]
+) -> dict:
+    """
+    Process the result of Transformers segmentation functions such as
+    `post_process_semantic_segmentation` and `post_process_instance_segmentation` (v5).
+
+    Args:
+        segmentation_result (dict): Dictionary containing segmentation results with keys
+            `segments_info` and `segmentation`.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+
+    Returns:
+        dict: Processed segmentation result including bounding boxes, masks, confidence
+              scores, class IDs, and data.
+    """
+    segments_info = segmentation_result["segments_info"]
+    scores = np.array([segment["score"] for segment in segments_info])
+    class_ids = np.array([segment["label_id"] for segment in segments_info])
+    segmentation_array = segmentation_result["segmentation"].cpu().detach().numpy()
+    masks = np.array(
+        [segmentation_array == segment["id"] for segment in segments_info]
+    ).astype(bool)
+    data = append_class_names_to_data(class_ids, id2label, {})
+
+    return dict(
+        xyxy=mask_to_xyxy(masks),
+        mask=masks,
+        confidence=scores,
+        class_id=class_ids,
+        data=data,
+    )
+
+
+def process_transformers_v4_panoptic_segmentation_result(
+    segmentation_result: dict, id2label: Optional[Dict[int, str]]
+) -> dict:
+    """
+    Process the result of the Transformers function `post_process_panoptic` (v4).
+
+    Args:
+        segmentation_result (dict): Dictionary containing segmentation results with keys
+            such as 'png_string' and 'segments_info'.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+
+    Returns:
+        dict: Processed segmentation result including bounding boxes, masks,
+              class IDs, and data.
+    """
+    segments_info = segmentation_result["segments_info"]
+    png_string = segmentation_result["png_string"]
+    class_ids = np.array([segment["category_id"] for segment in segments_info])
+    segmentation_array = png_string_to_segmentation_array(png_string=png_string)
+    masks = np.array(
+        [segmentation_array == segment["id"] for segment in segments_info]
+    ).astype(bool)
+    data = append_class_names_to_data(class_ids, id2label, {})
+
+    return dict(
+        xyxy=mask_to_xyxy(masks),
+        mask=masks,
+        class_id=class_ids,
+        data=data,
+    )
+
+
+def process_transformers_v5_panoptic_segmentation_result(
+    segmentation_array: np.ndarray, id2label: Optional[Dict[int, str]]
+) -> dict:
+    """
+    Process the result of the Transformers function
+    `post_process_panoptic_segmentation` (v5).
+
+    Args:
+        segmentation_array (np.ndarray): Segmentation array.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+
+    Returns:
+        dict: Processed segmentation result including bounding boxes, masks,
+              class IDs, and data.
+    """
+    class_ids = np.unique(segmentation_array)
+    masks = np.stack(
+        [segmentation_array == class_id for class_id in class_ids], axis=0
+    ).astype(bool)
+    data = append_class_names_to_data(class_ids, id2label, {})
+    return dict(xyxy=mask_to_xyxy(masks), mask=masks, class_id=class_ids, data=data)
+
+
+def png_string_to_segmentation_array(png_string: bytes) -> np.ndarray:
+    """
+    Convert a PNG byte string to a label mask array.
+
+    Args:
+        png_string (bytes): A byte string representing the PNG image.
+
+    Returns:
+        np.ndarray: A label mask array with shape (H, W), where H and W
+        are the height and width of the image. Each unique value in the array
+        represents a different object or category.
+    """
+    image = Image.open(io.BytesIO(png_string))
+    mask = np.array(image, dtype=np.uint8)
+    return mask[:, :, 0]
+
+
+def append_class_names_to_data(
+    class_ids: np.ndarray,
+    id2label: Optional[Dict[int, str]],
+    data: Optional[Dict[str, Any]] = None,
+) -> Dict[str, Any]:
+    """
+    Helper function to create or append to a data dictionary with class names if
+    available.
+
+    Args:
+        class_ids (np.ndarray): Array of class IDs.
+        id2label (Optional[Dict[int, str]]): A dictionary mapping class IDs to labels,
+            typically part of the `transformers` model configuration. If provided, the
+            resulting dictionary will include class names.
+        data (Optional[Dict[str, Any]]): An existing data dictionary to append to.
+
+    Returns:
+        Dict[str, Any]: Dictionary containing class names if id2label is provided.
+    """
+    if data is None:
+        data = {}
+
+    if id2label is not None:
+        class_names = np.array([id2label[class_id] for class_id in class_ids])
+        data[CLASS_NAME_DATA_FIELD] = class_names
+
+    return data

supervision/detection/utils.py

@@ -3,8 +3,10 @@ from typing import Dict, List, Optional, Tuple, Union
 
 import cv2
 import numpy as np
+import numpy.typing as npt
 
 from supervision.config import CLASS_NAME_DATA_FIELD
+from supervision.geometry.core import Vector
 
 MIN_POLYGON_POINT_COUNT = 3
 
@@ -56,7 +58,9 @@ def box_iou_batch(boxes_true: np.ndarray, boxes_detection: np.ndarray) -> np.nda
     bottom_right = np.minimum(boxes_true[:, None, 2:], boxes_detection[:, 2:])
 
     area_inter = np.prod(np.clip(bottom_right - top_left, a_min=0, a_max=None), 2)
-    return area_inter / (area_true[:, None] + area_detection - area_inter)
+    ious = area_inter / (area_true[:, None] + area_detection - area_inter)
+    ious = np.nan_to_num(ious)
+    return ious
 
 
 def _mask_iou_batch_split(
@@ -102,7 +106,7 @@ def mask_iou_batch(
     Args:
         masks_true (np.ndarray): 3D `np.ndarray` representing ground-truth masks.
         masks_detection (np.ndarray): 3D `np.ndarray` representing detection masks.
-        memory_limit (int, optional): memory limit in MB, default is 1024 * 5 MB (5GB).
+        memory_limit (int): memory limit in MB, default is 1024 * 5 MB (5GB).
 
     Returns:
         np.ndarray: Pairwise IoU of masks from `masks_true` and `masks_detection`.
@@ -136,144 +140,6 @@ def mask_iou_batch(
     return np.vstack(ious)
 
 
-def resize_masks(masks: np.ndarray, max_dimension: int = 640) -> np.ndarray:
-    """
-    Resize all masks in the array to have a maximum dimension of max_dimension,
-    maintaining aspect ratio.
-
-    Args:
-        masks (np.ndarray): 3D array of binary masks with shape (N, H, W).
-        max_dimension (int): The maximum dimension for the resized masks.
-
-    Returns:
-        np.ndarray: Array of resized masks.
-    """
-    max_height = np.max(masks.shape[1])
-    max_width = np.max(masks.shape[2])
-    scale = min(max_dimension / max_height, max_dimension / max_width)
-
-    new_height = int(scale * max_height)
-    new_width = int(scale * max_width)
-
-    x = np.linspace(0, max_width - 1, new_width).astype(int)
-    y = np.linspace(0, max_height - 1, new_height).astype(int)
-    xv, yv = np.meshgrid(x, y)
-
-    resized_masks = masks[:, yv, xv]
-
-    resized_masks = resized_masks.reshape(masks.shape[0], new_height, new_width)
-    return resized_masks
-
-
-def mask_non_max_suppression(
-    predictions: np.ndarray,
-    masks: np.ndarray,
-    iou_threshold: float = 0.5,
-    mask_dimension: int = 640,
-) -> np.ndarray:
-    """
-    Perform Non-Maximum Suppression (NMS) on segmentation predictions.
-
-    Args:
-        predictions (np.ndarray): A 2D array of object detection predictions in
-            the format of `(x_min, y_min, x_max, y_max, score)`
-            or `(x_min, y_min, x_max, y_max, score, class)`. Shape: `(N, 5)` or
-            `(N, 6)`, where N is the number of predictions.
-        masks (np.ndarray): A 3D array of binary masks corresponding to the predictions.
-            Shape: `(N, H, W)`, where N is the number of predictions, and H, W are the
-            dimensions of each mask.
-        iou_threshold (float, optional): The intersection-over-union threshold
-            to use for non-maximum suppression.
-        mask_dimension (int, optional): The dimension to which the masks should be
-            resized before computing IOU values. Defaults to 640.
-
-    Returns:
-        np.ndarray: A boolean array indicating which predictions to keep after
-            non-maximum suppression.
-
-    Raises:
-        AssertionError: If `iou_threshold` is not within the closed
-        range from `0` to `1`.
-    """
-    assert 0 <= iou_threshold <= 1, (
-        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
-        f"{iou_threshold} given."
-    )
-    rows, columns = predictions.shape
-
-    if columns == 5:
-        predictions = np.c_[predictions, np.zeros(rows)]
-
-    sort_index = predictions[:, 4].argsort()[::-1]
-    predictions = predictions[sort_index]
-    masks = masks[sort_index]
-    masks_resized = resize_masks(masks, mask_dimension)
-    ious = mask_iou_batch(masks_resized, masks_resized)
-    categories = predictions[:, 5]
-
-    keep = np.ones(rows, dtype=bool)
-    for i in range(rows):
-        if keep[i]:
-            condition = (ious[i] > iou_threshold) & (categories[i] == categories)
-            keep[i + 1 :] = np.where(condition[i + 1 :], False, keep[i + 1 :])
-
-    return keep[sort_index.argsort()]
-
-
-def box_non_max_suppression(
-    predictions: np.ndarray, iou_threshold: float = 0.5
-) -> np.ndarray:
-    """
-    Perform Non-Maximum Suppression (NMS) on object detection predictions.
-
-    Args:
-        predictions (np.ndarray): An array of object detection predictions in
-            the format of `(x_min, y_min, x_max, y_max, score)`
-            or `(x_min, y_min, x_max, y_max, score, class)`.
-        iou_threshold (float, optional): The intersection-over-union threshold
-            to use for non-maximum suppression.
-
-    Returns:
-        np.ndarray: A boolean array indicating which predictions to keep after n
-            on-maximum suppression.
-
-    Raises:
-        AssertionError: If `iou_threshold` is not within the
-            closed range from `0` to `1`.
-    """
-    assert 0 <= iou_threshold <= 1, (
-        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
-        f"{iou_threshold} given."
-    )
-    rows, columns = predictions.shape
-
-    # add column #5 - category filled with zeros for agnostic nms
-    if columns == 5:
-        predictions = np.c_[predictions, np.zeros(rows)]
-
-    # sort predictions column #4 - score
-    sort_index = np.flip(predictions[:, 4].argsort())
-    predictions = predictions[sort_index]
-
-    boxes = predictions[:, :4]
-    categories = predictions[:, 5]
-    ious = box_iou_batch(boxes, boxes)
-    ious = ious - np.eye(rows)
-
-    keep = np.ones(rows, dtype=bool)
-
-    for index, (iou, category) in enumerate(zip(ious, categories)):
-        if not keep[index]:
-            continue
-
-        # drop detections with iou > iou_threshold and
-        # same category as current detections
-        condition = (iou > iou_threshold) & (categories == category)
-        keep = keep & ~condition
-
-    return keep[sort_index.argsort()]
-
-
 def clip_boxes(xyxy: np.ndarray, resolution_wh: Tuple[int, int]) -> np.ndarray:
     """
     Clips bounding boxes coordinates to fit within the frame resolution.
@@ -289,6 +155,25 @@ def clip_boxes(xyxy: np.ndarray, resolution_wh: Tuple[int, int]) -> np.ndarray:
         np.ndarray: A numpy array of shape `(N, 4)` where each row
             corresponds to a bounding box with coordinates clipped to fit
             within the frame resolution.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xyxy = np.array([
+            [10, 20, 300, 200],
+            [15, 25, 350, 450],
+            [-10, -20, 30, 40]
+        ])
+
+        sv.clip_boxes(xyxy=xyxy, resolution_wh=(320, 240))
+        # array([
+        #     [ 10,  20, 300, 200],
+        #     [ 15,  25, 320, 240],
+        #     [  0,   0,  30,  40]
+        # ])
+        ```
     """
     result = np.copy(xyxy)
     width, height = resolution_wh
@@ -297,10 +182,124 @@ def clip_boxes(xyxy: np.ndarray, resolution_wh: Tuple[int, int]) -> np.ndarray:
     return result
 
 
-def xywh_to_xyxy(boxes_xywh: np.ndarray) -> np.ndarray:
-    xyxy = boxes_xywh.copy()
-    xyxy[:, 2] = boxes_xywh[:, 0] + boxes_xywh[:, 2]
-    xyxy[:, 3] = boxes_xywh[:, 1] + boxes_xywh[:, 3]
+def pad_boxes(xyxy: np.ndarray, px: int, py: Optional[int] = None) -> np.ndarray:
+    """
+    Pads bounding boxes coordinates with a constant padding.
+
+    Args:
+        xyxy (np.ndarray): A numpy array of shape `(N, 4)` where each
+            row corresponds to a bounding box in the format
+            `(x_min, y_min, x_max, y_max)`.
+        px (int): The padding value to be added to both the left and right sides of
+            each bounding box.
+        py (Optional[int]): The padding value to be added to both the top and bottom
+            sides of each bounding box. If not provided, `px` will be used for both
+            dimensions.
+
+    Returns:
+        np.ndarray: A numpy array of shape `(N, 4)` where each row corresponds to a
+            bounding box with coordinates padded according to the provided padding
+            values.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xyxy = np.array([
+            [10, 20, 30, 40],
+            [15, 25, 35, 45]
+        ])
+
+        sv.pad_boxes(xyxy=xyxy, px=5, py=10)
+        # array([
+        #     [ 5, 10, 35, 50],
+        #     [10, 15, 40, 55]
+        # ])
+        ```
+    """
+    if py is None:
+        py = px
+
+    result = xyxy.copy()
+    result[:, [0, 1]] -= [px, py]
+    result[:, [2, 3]] += [px, py]
+
+    return result
+
+
+def xywh_to_xyxy(xywh: np.ndarray) -> np.ndarray:
+    """
+    Converts bounding box coordinates from `(x, y, width, height)`
+    format to `(x_min, y_min, x_max, y_max)` format.
+
+    Args:
+        xywh (np.ndarray): A numpy array of shape `(N, 4)` where each row
+            corresponds to a bounding box in the format `(x, y, width, height)`.
+
+    Returns:
+        np.ndarray: A numpy array of shape `(N, 4)` where each row corresponds
+            to a bounding box in the format `(x_min, y_min, x_max, y_max)`.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xywh = np.array([
+            [10, 20, 30, 40],
+            [15, 25, 35, 45]
+        ])
+
+        sv.xywh_to_xyxy(xywh=xywh)
+        # array([
+        #     [10, 20, 40, 60],
+        #     [15, 25, 50, 70]
+        # ])
+        ```
+    """
+    xyxy = xywh.copy()
+    xyxy[:, 2] = xywh[:, 0] + xywh[:, 2]
+    xyxy[:, 3] = xywh[:, 1] + xywh[:, 3]
+    return xyxy
+
+
+def xcycwh_to_xyxy(xcycwh: np.ndarray) -> np.ndarray:
+    """
+    Converts bounding box coordinates from `(center_x, center_y, width, height)`
+    format to `(x_min, y_min, x_max, y_max)` format.
+
+    Args:
+        xcycwh (np.ndarray): A numpy array of shape `(N, 4)` where each row
+            corresponds to a bounding box in the format `(center_x, center_y, width,
+            height)`.
+
+    Returns:
+        np.ndarray: A numpy array of shape `(N, 4)` where each row corresponds
+            to a bounding box in the format `(x_min, y_min, x_max, y_max)`.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xcycwh = np.array([
+            [50, 50, 20, 30],
+            [30, 40, 10, 15]
+        ])
+
+        sv.xcycwh_to_xyxy(xcycwh=xcycwh)
+        # array([
+        #     [40, 35, 60, 65],
+        #     [25, 32.5, 35, 47.5]
+        # ])
+        ```
+    """
+    xyxy = xcycwh.copy()
+    xyxy[:, 0] = xcycwh[:, 0] - xcycwh[:, 2] / 2
+    xyxy[:, 1] = xcycwh[:, 1] - xcycwh[:, 3] / 2
+    xyxy[:, 2] = xcycwh[:, 0] + xcycwh[:, 2] / 2
+    xyxy[:, 3] = xcycwh[:, 1] + xcycwh[:, 3] / 2
     return xyxy
 
 
@@ -317,7 +316,7 @@ def mask_to_xyxy(masks: np.ndarray) -> np.ndarray:
             `(x_min, y_min, x_max, y_max)` for each mask
     """
     n = masks.shape[0]
-    bboxes = np.zeros((n, 4), dtype=int)
+    xyxy = np.zeros((n, 4), dtype=int)
 
     for i, mask in enumerate(masks):
         rows, cols = np.where(mask)
@@ -325,9 +324,9 @@ def mask_to_xyxy(masks: np.ndarray) -> np.ndarray:
         if len(rows) > 0 and len(cols) > 0:
             x_min, x_max = np.min(cols), np.max(cols)
             y_min, y_max = np.min(rows), np.max(rows)
-            bboxes[i, :] = [x_min, y_min, x_max, y_max]
+            xyxy[i, :] = [x_min, y_min, x_max, y_max]
 
-    return bboxes
+    return xyxy
 
 
 def mask_to_polygons(mask: np.ndarray) -> List[np.ndarray]:
@@ -500,7 +499,7 @@ def process_roboflow_result(
     np.ndarray,
     Optional[np.ndarray],
     Optional[np.ndarray],
-    Dict[str, List[np.ndarray]],
+    Dict[str, Union[List[np.ndarray], np.ndarray]],
 ]:
     if not roboflow_result["predictions"]:
         return (
@@ -563,59 +562,158 @@ def process_roboflow_result(
     return xyxy, confidence, class_id, masks, tracker_id, data
 
 
-def move_boxes(xyxy: np.ndarray, offset: np.ndarray) -> np.ndarray:
+def move_boxes(
+    xyxy: npt.NDArray[np.float64], offset: npt.NDArray[np.int32]
+) -> npt.NDArray[np.float64]:
     """
     Parameters:
-        xyxy (np.ndarray): An array of shape `(n, 4)` containing the bounding boxes
-            coordinates in format `[x1, y1, x2, y2]`
+        xyxy (npt.NDArray[np.float64]): An array of shape `(n, 4)` containing the
+            bounding boxes coordinates in format `[x1, y1, x2, y2]`
         offset (np.array): An array of shape `(2,)` containing offset values in format
             is `[dx, dy]`.
 
     Returns:
-        np.ndarray: Repositioned bounding boxes.
+        npt.NDArray[np.float64]: Repositioned bounding boxes.
 
-    Example:
+    Examples:
         ```python
         import numpy as np
         import supervision as sv
 
-        boxes = np.array([[10, 10, 20, 20], [30, 30, 40, 40]])
+        xyxy = np.array([
+            [10, 10, 20, 20],
+            [30, 30, 40, 40]
+        ])
         offset = np.array([5, 5])
-        moved_box = sv.move_boxes(boxes, offset)
-        print(moved_box)
-        # np.array([
+
+        sv.move_boxes(xyxy=xyxy, offset=offset)
+        # array([
         #    [15, 15, 25, 25],
-        #     [35, 35, 45, 45]
+        #    [35, 35, 45, 45]
         # ])
         ```
     """
     return xyxy + np.hstack([offset, offset])
 
 
-def scale_boxes(xyxy: np.ndarray, factor: float) -> np.ndarray:
+def move_oriented_boxes(
+    xyxyxyxy: npt.NDArray[np.float64], offset: npt.NDArray[np.int32]
+) -> npt.NDArray[np.float64]:
+    """
+    Parameters:
+    xyxyxyxy (npt.NDArray[np.float64]): An array of shape `(n, 4, 2)` containing the
+    oriented bounding boxes coordinates in format
+    `[[x1, y1], [x2, y2], [x3, y3], [x3, y3]]`
+    offset (np.array): An array of shape `(2,)` containing offset values in format
+        is `[dx, dy]`.
+
+    Returns:
+    npt.NDArray[np.float64]: Repositioned bounding boxes.
+
+    Examples:
+    ```python
+    import numpy as np
+    import supervision as sv
+
+    xyxyxyxy = np.array([
+        [
+            [20, 10],
+            [10, 20],
+            [20, 30],
+            [30, 20]
+        ],
+        [
+            [30 ,30],
+            [20, 40],
+            [30, 50],
+            [40, 40]
+        ]
+    ])
+    offset = np.array([5, 5])
+
+    sv.move_oriented_boxes(xyxy=xyxy, offset=offset)
+    # array([
+    #     [
+    #         [25, 15],
+    #         [15, 25],
+    #         [25, 35],
+    #         [35, 25]
+    #     ],
+    #     [
+    #         [35, 35],
+    #         [25, 45],
+    #         [35, 55],
+    #         [45, 45]
+    #     ]
+    # ])
+    ```
+    """
+    return xyxyxyxy + offset
+
+
+def move_masks(
+    masks: npt.NDArray[np.bool_],
+    offset: npt.NDArray[np.int32],
+    resolution_wh: Tuple[int, int],
+) -> npt.NDArray[np.bool_]:
+    """
+    Offset the masks in an array by the specified (x, y) amount.
+
+    Args:
+        masks (npt.NDArray[np.bool_]): A 3D array of binary masks corresponding to the
+            predictions. Shape: `(N, H, W)`, where N is the number of predictions, and
+            H, W are the dimensions of each mask.
+        offset (npt.NDArray[np.int32]): An array of shape `(2,)` containing non-negative
+            int values `[dx, dy]`.
+        resolution_wh (Tuple[int, int]): The width and height of the desired mask
+            resolution.
+
+    Returns:
+        (npt.NDArray[np.bool_]) repositioned masks, optionally padded to the specified
+            shape.
+    """
+
+    if offset[0] < 0 or offset[1] < 0:
+        raise ValueError(f"Offset values must be non-negative integers. Got: {offset}")
+
+    mask_array = np.full((masks.shape[0], resolution_wh[1], resolution_wh[0]), False)
+    mask_array[
+        :,
+        offset[1] : masks.shape[1] + offset[1],
+        offset[0] : masks.shape[2] + offset[0],
+    ] = masks
+
+    return mask_array
+
+
+def scale_boxes(
+    xyxy: npt.NDArray[np.float64], factor: float
+) -> npt.NDArray[np.float64]:
     """
     Scale the dimensions of bounding boxes.
 
     Parameters:
-        xyxy (np.ndarray): An array of shape `(n, 4)` containing the bounding boxes
-            coordinates in format `[x1, y1, x2, y2]`
+        xyxy (npt.NDArray[np.float64]): An array of shape `(n, 4)` containing the
+            bounding boxes coordinates in format `[x1, y1, x2, y2]`
         factor (float): A float value representing the factor by which the box
             dimensions are scaled. A factor greater than 1 enlarges the boxes, while a
             factor less than 1 shrinks them.
 
     Returns:
-        np.ndarray: Scaled bounding boxes.
+        npt.NDArray[np.float64]: Scaled bounding boxes.
 
-    Example:
+    Examples:
         ```python
         import numpy as np
         import supervision as sv
 
-        boxes = np.array([[10, 10, 20, 20], [30, 30, 40, 40]])
-        factor = 1.5
-        scaled_bb = sv.scale_boxes(boxes, factor)
-        print(scaled_bb)
-        # np.array([
+        xyxy = np.array([
+            [10, 10, 20, 20],
+            [30, 30, 40, 40]
+        ])
+
+        sv.scale_boxes(xyxy=xyxy, factor=1.5)
+        # array([
         #    [ 7.5,  7.5, 22.5, 22.5],
         #    [27.5, 27.5, 42.5, 42.5]
         # ])
@@ -672,17 +770,19 @@ def is_data_equal(data_a: Dict[str, np.ndarray], data_b: Dict[str, np.ndarray])
 
 
 def merge_data(
-    data_list: List[Dict[str, Union[np.ndarray, List]]],
-) -> Dict[str, Union[np.ndarray, List]]:
+    data_list: List[Dict[str, Union[npt.NDArray[np.generic], List]]],
+) -> Dict[str, Union[npt.NDArray[np.generic], List]]:
     """
     Merges the data payloads of a list of Detections instances.
 
     Args:
-        data_list: The data payloads of the instances.
+        data_list: The data payloads of the Detections instances. Each data payload
+            is a dictionary with the same keys, and the values are either lists or
+            npt.NDArray[np.generic].
 
     Returns:
         A single data payload containing the merged data, preserving the original data
-            types (list or np.ndarray).
+            types (list or npt.NDArray[np.generic]).
 
     Raises:
         ValueError: If data values within a single object have different lengths or if
@@ -703,9 +803,8 @@ def merge_data(
             )
 
     merged_data = {key: [] for key in all_keys_sets[0]}
-
     for data in data_list:
-        for key in merged_data:
+        for key in data:
             merged_data[key].append(data[key])
 
     for key in merged_data:
@@ -766,3 +865,138 @@ def get_data_item(
             raise TypeError(f"Unsupported data type for key '{key}': {type(value)}")
 
     return subset_data
+
+
+def contains_holes(mask: npt.NDArray[np.bool_]) -> bool:
+    """
+    Checks if the binary mask contains holes (background pixels fully enclosed by
+    foreground pixels).
+
+    Args:
+        mask (npt.NDArray[np.bool_]): 2D binary mask where `True` indicates foreground
+            object and `False` indicates background.
+
+    Returns:
+        True if holes are detected, False otherwise.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([
+            [0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 0],
+            [0, 1, 0, 1, 0],
+            [0, 1, 1, 1, 0],
+            [0, 0, 0, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_holes(mask=mask)
+        # True
+
+        mask = np.array([
+            [0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 0],
+            [0, 1, 1, 1, 0],
+            [0, 1, 1, 1, 0],
+            [0, 0, 0, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_holes(mask=mask)
+        # False
+        ```
+
+    ![contains_holes](https://media.roboflow.com/supervision-docs/contains-holes.png){ align=center width="800" }
+    """  # noqa E501 // docs
+    mask_uint8 = mask.astype(np.uint8)
+    _, hierarchy = cv2.findContours(mask_uint8, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_SIMPLE)
+
+    if hierarchy is not None:
+        parent_contour_index = 3
+        for h in hierarchy[0]:
+            if h[parent_contour_index] != -1:
+                return True
+    return False
+
+
+def contains_multiple_segments(
+    mask: npt.NDArray[np.bool_], connectivity: int = 4
+) -> bool:
+    """
+    Checks if the binary mask contains multiple unconnected foreground segments.
+
+    Args:
+        mask (npt.NDArray[np.bool_]): 2D binary mask where `True` indicates foreground
+            object and `False` indicates background.
+        connectivity (int) : Default: 4 is 4-way connectivity, which means that
+            foreground pixels are the part of the same segment/component
+            if their edges touch.
+            Alternatively: 8 for 8-way connectivity, when foreground pixels are
+            connected by their edges or corners touch.
+
+    Returns:
+        True when the mask contains multiple not connected components, False otherwise.
+
+    Raises:
+        ValueError: If connectivity(int) parameter value is not 4 or 8.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([
+            [0, 0, 0, 0, 0, 0],
+            [0, 1, 1, 0, 1, 1],
+            [0, 1, 1, 0, 1, 1],
+            [0, 0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 0, 0],
+            [0, 1, 1, 1, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_multiple_segments(mask=mask, connectivity=4)
+        # True
+
+        mask = np.array([
+            [0, 0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 1, 1],
+            [0, 1, 1, 1, 1, 1],
+            [0, 1, 1, 1, 1, 1],
+            [0, 1, 1, 1, 1, 1],
+            [0, 0, 0, 0, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_multiple_segments(mask=mask, connectivity=4)
+        # False
+        ```
+
+    ![contains_multiple_segments](https://media.roboflow.com/supervision-docs/contains-multiple-segments.png){ align=center width="800" }
+    """  # noqa E501 // docs
+    if connectivity != 4 and connectivity != 8:
+        raise ValueError(
+            "Incorrect connectivity value. Possible connectivity values: 4 or 8."
+        )
+    mask_uint8 = mask.astype(np.uint8)
+    labels = np.zeros_like(mask_uint8, dtype=np.int32)
+    number_of_labels, _ = cv2.connectedComponents(
+        mask_uint8, labels, connectivity=connectivity
+    )
+    return number_of_labels > 2
+
+
+def cross_product(anchors: np.ndarray, vector: Vector) -> np.ndarray:
+    """
+    Get array of cross products of each anchor with a vector.
+    Args:
+        anchors: Array of anchors of shape (number of anchors, detections, 2)
+        vector: Vector to calculate cross product with
+
+    Returns:
+        Array of cross products of shape (number of anchors, detections)
+    """
+    vector_at_zero = np.array(
+        [vector.end.x - vector.start.x, vector.end.y - vector.start.y]
+    )
+    vector_start = np.array([vector.start.x, vector.start.y])
+    return np.cross(vector_at_zero, anchors - vector_start)

supervision/draw/color.py

@@ -5,7 +5,7 @@ from typing import List, Tuple, Union
 
 import matplotlib.pyplot as plt
 
-from supervision.utils.internal import classproperty, deprecated
+from supervision.utils.internal import classproperty
 
 DEFAULT_COLOR_PALETTE = [
     "A351FB",
@@ -255,46 +255,6 @@ class Color:
     def ROBOFLOW(cls) -> Color:
         return Color.from_hex("#A351FB")
 
-    @classmethod
-    @deprecated(
-        "`Color.white()` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Color.WHITE` instead."
-    )
-    def white(cls) -> Color:
-        return Color.from_hex(color_hex="#ffffff")
-
-    @classmethod
-    @deprecated(
-        "`Color.black()` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Color.BLACK` instead."
-    )
-    def black(cls) -> Color:
-        return Color.from_hex(color_hex="#000000")
-
-    @classmethod
-    @deprecated(
-        "`Color.red()` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Color.RED` instead."
-    )
-    def red(cls) -> Color:
-        return Color.from_hex(color_hex="#ff0000")
-
-    @classmethod
-    @deprecated(
-        "`Color.green()` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Color.GREEN` instead."
-    )
-    def green(cls) -> Color:
-        return Color.from_hex(color_hex="#00ff00")
-
-    @classmethod
-    @deprecated(
-        "`Color.blue()` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Color.BLUE` instead."
-    )
-    def blue(cls) -> Color:
-        return Color.from_hex(color_hex="#0000ff")
-
 
 @dataclass
 class ColorPalette:
@@ -346,33 +306,6 @@ class ColorPalette:
     def LEGACY(cls) -> ColorPalette:
         return ColorPalette.from_hex(color_hex_list=LEGACY_COLOR_PALETTE)
 
-    @classmethod
-    @deprecated(
-        "`ColorPalette.default()` is deprecated and will be removed in "
-        "`supervision-0.22.0`. Use `Color.DEFAULT` instead."
-    )
-    def default(cls) -> ColorPalette:
-        """
-        !!! failure "Deprecated"
-
-            `ColorPalette.default()` is deprecated and will be removed in
-            `supervision-0.22.0`. Use `Color.DEFAULT` instead.
-
-        Returns a default color palette.
-
-        Returns:
-            ColorPalette: A ColorPalette instance with default colors.
-
-        Example:
-            ```python
-            import supervision as sv
-
-            sv.ColorPalette.default()
-            # ColorPalette(colors=[Color(r=255, g=64, b=64), Color(r=255, g=161, b=160), ...])
-            ```
-        """  # noqa: E501 // docs
-        return ColorPalette.from_hex(color_hex_list=DEFAULT_COLOR_PALETTE)
-
     @classmethod
     def from_hex(cls, color_hex_list: List[str]) -> ColorPalette:
         """
@@ -419,11 +352,15 @@ class ColorPalette:
         supervision-annotator-examples/visualized_color_palette.png)
         """  # noqa: E501 // docs
         mpl_palette = plt.get_cmap(palette_name, color_count)
-        colors = [
-            Color(int(r * 255), int(g * 255), int(b * 255))
-            for r, g, b, _ in mpl_palette.colors
-        ]
-        return cls(colors)
+
+        if hasattr(mpl_palette, "colors"):
+            colors = mpl_palette.colors
+        else:
+            colors = [mpl_palette(i / (color_count - 1)) for i in range(color_count)]
+
+        return cls(
+            [Color(int(r * 255), int(g * 255), int(b * 255)) for r, g, b, _ in colors]
+        )
 
     def by_idx(self, idx: int) -> Color:
         """

supervision/draw/utils.py

@@ -81,6 +81,58 @@ def draw_filled_rectangle(scene: np.ndarray, rect: Rect, color: Color) -> np.nda
     return scene
 
 
+def draw_rounded_rectangle(
+    scene: np.ndarray,
+    rect: Rect,
+    color: Color,
+    border_radius: int,
+) -> np.ndarray:
+    """
+    Draws a rounded rectangle on an image.
+
+    Parameters:
+        scene (np.ndarray): The image on which the rounded rectangle will be drawn.
+        rect (Rect): The rectangle to be drawn.
+        color (Color): The color of the rounded rectangle.
+        border_radius (int): The radius of the corner rounding.
+
+    Returns:
+        np.ndarray: The image with the rounded rectangle drawn on it.
+    """
+    x1, y1, x2, y2 = rect.as_xyxy_int_tuple()
+    width, height = x2 - x1, y2 - y1
+    border_radius = min(border_radius, min(width, height) // 2)
+
+    rectangle_coordinates = [
+        ((x1 + border_radius, y1), (x2 - border_radius, y2)),
+        ((x1, y1 + border_radius), (x2, y2 - border_radius)),
+    ]
+    circle_centers = [
+        (x1 + border_radius, y1 + border_radius),
+        (x2 - border_radius, y1 + border_radius),
+        (x1 + border_radius, y2 - border_radius),
+        (x2 - border_radius, y2 - border_radius),
+    ]
+
+    for coordinates in rectangle_coordinates:
+        cv2.rectangle(
+            img=scene,
+            pt1=coordinates[0],
+            pt2=coordinates[1],
+            color=color.as_bgr(),
+            thickness=-1,
+        )
+    for center in circle_centers:
+        cv2.circle(
+            img=scene,
+            center=center,
+            radius=border_radius,
+            color=color.as_bgr(),
+            thickness=-1,
+        )
+    return scene
+
+
 def draw_polygon(
     scene: np.ndarray, polygon: np.ndarray, color: Color, thickness: int = 2
 ) -> np.ndarray:
@@ -90,7 +142,7 @@ def draw_polygon(
         scene (np.ndarray): The scene to draw the polygon on.
         polygon (np.ndarray): The polygon to be drawn, given as a list of vertices.
         color (Color): The color of the polygon.
-        thickness (int, optional): The thickness of the polygon lines, by default 2.
+        thickness (int): The thickness of the polygon lines, by default 2.
 
     Returns:
         np.ndarray: The scene with the polygon drawn on it.
@@ -120,14 +172,14 @@ def draw_text(
         text (str): The text to be drawn.
         text_anchor (Point): The anchor point for the text, represented as a
             Point object with x and y attributes.
-        text_color (Color, optional): The color of the text. Defaults to black.
-        text_scale (float, optional): The scale of the text. Defaults to 0.5.
-        text_thickness (int, optional): The thickness of the text. Defaults to 1.
-        text_padding (int, optional): The amount of padding to add around the text
+        text_color (Color): The color of the text. Defaults to black.
+        text_scale (float): The scale of the text. Defaults to 0.5.
+        text_thickness (int): The thickness of the text. Defaults to 1.
+        text_padding (int): The amount of padding to add around the text
             when drawing a rectangle in the background. Defaults to 10.
-        text_font (int, optional): The font to use for the text.
+        text_font (int): The font to use for the text.
             Defaults to cv2.FONT_HERSHEY_SIMPLEX.
-        background_color (Color, optional): The color of the background rectangle,
+        background_color (Optional[Color]): The color of the background rectangle,
             if one is to be drawn. Defaults to None.
 
     Returns:
@@ -243,8 +295,8 @@ def calculate_optimal_text_scale(resolution_wh: Tuple[int, int]) -> float:
     Calculate font scale based on the resolution of an image.
 
     Parameters:
-         resolution_wh (Tuple[int, int]): A tuple representing the width and height
-             of the image.
+        resolution_wh (Tuple[int, int]): A tuple representing the width and height
+            of the image.
 
     Returns:
          float: The calculated font scale factor.

supervision/geometry/core.py

@@ -98,6 +98,11 @@ class Rect:
     width: float
     height: float
 
+    @classmethod
+    def from_xyxy(cls, xyxy: Tuple[float, float, float, float]) -> Rect:
+        x1, y1, x2, y2 = xyxy
+        return cls(x=x1, y=y1, width=x2 - x1, height=y2 - y1)
+
     @property
     def top_left(self) -> Point:
         return Point(x=self.x, y=self.y)
@@ -113,3 +118,11 @@ class Rect:
             width=self.width + 2 * padding,
             height=self.height + 2 * padding,
         )
+
+    def as_xyxy_int_tuple(self) -> Tuple[int, int, int, int]:
+        return (
+            int(self.x),
+            int(self.y),
+            int(self.x + self.width),
+            int(self.y + self.height),
+        )

supervision/keypoint/annotators.py

@@ -1,15 +1,17 @@
 from abc import ABC, abstractmethod
 from logging import warn
-from typing import List, Optional, Tuple
+from typing import List, Optional, Tuple, Union
 
 import cv2
 import numpy as np
 
+from supervision import Rect, pad_boxes
 from supervision.annotators.base import ImageType
 from supervision.draw.color import Color
+from supervision.draw.utils import draw_rounded_rectangle
 from supervision.keypoint.core import KeyPoints
 from supervision.keypoint.skeletons import SKELETONS_BY_VERTEX_COUNT
-from supervision.utils.conversion import convert_for_annotation_method
+from supervision.utils.conversion import ensure_cv2_image_for_annotation
 
 
 class BaseKeyPointAnnotator(ABC):
@@ -32,22 +34,22 @@ class VertexAnnotator(BaseKeyPointAnnotator):
     ) -> None:
         """
         Args:
-            color (Color, optional): The color to use for annotating key points.
-            radius (int, optional): The radius of the circles used to represent the key
+            color (Color): The color to use for annotating key points.
+            radius (int): The radius of the circles used to represent the key
                 points.
         """
         self.color = color
         self.radius = radius
 
-    @convert_for_annotation_method
+    @ensure_cv2_image_for_annotation
     def annotate(self, scene: ImageType, key_points: KeyPoints) -> ImageType:
         """
         Annotates the given scene with skeleton vertices based on the provided key
         points. It draws circles at each key point location.
 
         Args:
-            scene (ImageType): The image where bounding boxes will be drawn. `ImageType`
-                is a flexible type, accepting either `numpy.ndarray` or
+            scene (ImageType): The image where skeleton vertices will be drawn.
+                `ImageType` is a flexible type, accepting either `numpy.ndarray` or
                 `PIL.Image.Image`.
             key_points (KeyPoints): A collection of key points where each key point
                 consists of x and y coordinates.
@@ -63,7 +65,10 @@ class VertexAnnotator(BaseKeyPointAnnotator):
             image = ...
             key_points = sv.KeyPoints(...)
 
-            vertex_annotator = sv.VertexAnnotator(color=sv.Color.GREEN, radius=10)
+            vertex_annotator = sv.VertexAnnotator(
+                color=sv.Color.GREEN,
+                radius=10
+            )
             annotated_frame = vertex_annotator.annotate(
                 scene=image.copy(),
                 key_points=key_points
@@ -73,6 +78,7 @@ class VertexAnnotator(BaseKeyPointAnnotator):
         ![vertex-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/vertex-annotator-example.png)
         """
+        assert isinstance(scene, np.ndarray)
         if len(key_points) == 0:
             return scene
 
@@ -103,8 +109,8 @@ class EdgeAnnotator(BaseKeyPointAnnotator):
     ) -> None:
         """
         Args:
-            color (Color, optional): The color to use for the edges.
-            thickness (int, optional): The thickness of the edges.
+            color (Color): The color to use for the edges.
+            thickness (int): The thickness of the edges.
             edges (Optional[List[Tuple[int, int]]]): The edges to draw.
                 If set to `None`, will attempt to select automatically.
         """
@@ -112,14 +118,14 @@ class EdgeAnnotator(BaseKeyPointAnnotator):
         self.thickness = thickness
         self.edges = edges
 
-    @convert_for_annotation_method
+    @ensure_cv2_image_for_annotation
     def annotate(self, scene: ImageType, key_points: KeyPoints) -> ImageType:
         """
         Annotates the given scene by drawing lines between specified key points to form
         edges.
 
         Args:
-            scene (ImageType): The image where bounding boxes will be drawn. `ImageType`
+            scene (ImageType): The image where skeleton edges will be drawn. `ImageType`
                 is a flexible type, accepting either `numpy.ndarray` or
                 `PIL.Image.Image`.
             key_points (KeyPoints): A collection of key points where each key point
@@ -137,7 +143,10 @@ class EdgeAnnotator(BaseKeyPointAnnotator):
             image = ...
             key_points = sv.KeyPoints(...)
 
-            edge_annotator = sv.EdgeAnnotator(color=sv.Color.GREEN, thickness=5)
+            edge_annotator = sv.EdgeAnnotator(
+                color=sv.Color.GREEN,
+                thickness=5
+            )
             annotated_frame = edge_annotator.annotate(
                 scene=image.copy(),
                 key_points=key_points
@@ -147,6 +156,7 @@ class EdgeAnnotator(BaseKeyPointAnnotator):
         ![edge-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/edge-annotator-example.png)
         """
+        assert isinstance(scene, np.ndarray)
         if len(key_points) == 0:
             return scene
 
@@ -175,3 +185,251 @@ class EdgeAnnotator(BaseKeyPointAnnotator):
                 )
 
         return scene
+
+
+class VertexLabelAnnotator:
+    """
+    A class that draws labels of skeleton vertices on images. It uses specified key
+    points to determine the locations where the vertices should be drawn.
+    """
+
+    def __init__(
+        self,
+        color: Union[Color, List[Color]] = Color.ROBOFLOW,
+        text_color: Union[Color, List[Color]] = Color.WHITE,
+        text_scale: float = 0.5,
+        text_thickness: int = 1,
+        text_padding: int = 10,
+        border_radius: int = 0,
+    ):
+        """
+        Args:
+            color (Union[Color, List[Color]]): The color to use for each
+                keypoint label. If a list is provided, the colors will be used in order
+                for each keypoint.
+            text_color (Union[Color, List[Color]]): The color to use
+                for the labels. If a list is provided, the colors will be used in order
+                for each keypoint.
+            text_scale (float): The scale of the text.
+            text_thickness (int): The thickness of the text.
+            text_padding (int): The padding around the text.
+            border_radius (int): The radius of the rounded corners of the
+                boxes. Set to a high value to produce circles.
+        """
+        self.border_radius: int = border_radius
+        self.color: Union[Color, List[Color]] = color
+        self.text_color: Union[Color, List[Color]] = text_color
+        self.text_scale: float = text_scale
+        self.text_thickness: int = text_thickness
+        self.text_padding: int = text_padding
+
+    def annotate(
+        self,
+        scene: ImageType,
+        key_points: KeyPoints,
+        labels: Optional[List[str]] = None,
+    ) -> ImageType:
+        """
+        A class that draws labels of skeleton vertices on images. It uses specified key
+            points to determine the locations where the vertices should be drawn.
+
+        Args:
+            scene (ImageType): The image where vertex labels will be drawn. `ImageType`
+                is a flexible type, accepting either `numpy.ndarray` or
+                `PIL.Image.Image`.
+            key_points (KeyPoints): A collection of key points where each key point
+                consists of x and y coordinates.
+            labels (Optional[List[str]]): A list of labels to be displayed on the
+                annotated image. If not provided, keypoint indices will be used.
+
+        Returns:
+            The annotated image, matching the type of `scene` (`numpy.ndarray`
+                or `PIL.Image.Image`)
+
+        Example:
+            ```python
+            import supervision as sv
+
+            image = ...
+            key_points = sv.KeyPoints(...)
+
+            vertex_label_annotator = sv.VertexLabelAnnotator(
+                color=sv.Color.GREEN,
+                text_color=sv.Color.BLACK,
+                border_radius=5
+            )
+            annotated_frame = vertex_label_annotator.annotate(
+                scene=image.copy(),
+                key_points=key_points
+            )
+            ```
+
+        ![vertex-label-annotator-example](https://media.roboflow.com/
+        supervision-annotator-examples/vertex-label-annotator-example.png)
+
+        !!! tip
+
+            `VertexLabelAnnotator` allows to customize the color of each keypoint label
+            values.
+
+        Example:
+            ```python
+            import supervision as sv
+
+            image = ...
+            key_points = sv.KeyPoints(...)
+
+            LABELS = [
+                "nose", "left eye", "right eye", "left ear",
+                "right ear", "left shoulder", "right shoulder", "left elbow",
+                "right elbow", "left wrist", "right wrist", "left hip",
+                "right hip", "left knee", "right knee", "left ankle",
+                "right ankle"
+            ]
+
+            COLORS = [
+                "#FF6347", "#FF6347", "#FF6347", "#FF6347",
+                "#FF6347", "#FF1493", "#00FF00", "#FF1493",
+                "#00FF00", "#FF1493", "#00FF00", "#FFD700",
+                "#00BFFF", "#FFD700", "#00BFFF", "#FFD700",
+                "#00BFFF"
+            ]
+            COLORS = [sv.Color.from_hex(color_hex=c) for c in COLORS]
+
+            vertex_label_annotator = sv.VertexLabelAnnotator(
+                color=COLORS,
+                text_color=sv.Color.BLACK,
+                border_radius=5
+            )
+            annotated_frame = vertex_label_annotator.annotate(
+                scene=image.copy(),
+                key_points=key_points,
+                labels=labels
+            )
+            ```
+        ![vertex-label-annotator-custom-example](https://media.roboflow.com/
+        supervision-annotator-examples/vertex-label-annotator-custom-example.png)
+        """
+        assert isinstance(scene, np.ndarray)
+        font = cv2.FONT_HERSHEY_SIMPLEX
+
+        skeletons_count, points_count, _ = key_points.xy.shape
+        if skeletons_count == 0:
+            return scene
+
+        anchors = key_points.xy.reshape(points_count * skeletons_count, 2).astype(int)
+        mask = np.all(anchors != 0, axis=1)
+
+        if not np.any(mask):
+            return scene
+
+        colors = self.preprocess_and_validate_colors(
+            colors=self.color,
+            points_count=points_count,
+            skeletons_count=skeletons_count,
+        )
+
+        text_colors = self.preprocess_and_validate_colors(
+            colors=self.text_color,
+            points_count=points_count,
+            skeletons_count=skeletons_count,
+        )
+
+        labels = self.preprocess_and_validate_labels(
+            labels=labels, points_count=points_count, skeletons_count=skeletons_count
+        )
+
+        anchors = anchors[mask]
+        colors = colors[mask]
+        text_colors = text_colors[mask]
+        labels = labels[mask]
+
+        xyxy = np.array(
+            [
+                self.get_text_bounding_box(
+                    text=label,
+                    font=font,
+                    text_scale=self.text_scale,
+                    text_thickness=self.text_thickness,
+                    center_coordinates=tuple(anchor),
+                )
+                for anchor, label in zip(anchors, labels)
+            ]
+        )
+
+        xyxy_padded = pad_boxes(xyxy=xyxy, px=self.text_padding)
+
+        for text, color, text_color, box, box_padded in zip(
+            labels, colors, text_colors, xyxy, xyxy_padded
+        ):
+            draw_rounded_rectangle(
+                scene=scene,
+                rect=Rect.from_xyxy(box_padded),
+                color=color,
+                border_radius=self.border_radius,
+            )
+            cv2.putText(
+                img=scene,
+                text=text,
+                org=(box[0], box[3]),
+                fontFace=font,
+                fontScale=self.text_scale,
+                color=text_color.as_bgr(),
+                thickness=self.text_thickness,
+                lineType=cv2.LINE_AA,
+            )
+
+        return scene
+
+    @staticmethod
+    def get_text_bounding_box(
+        text: str,
+        font: int,
+        text_scale: float,
+        text_thickness: int,
+        center_coordinates: Tuple[int, int],
+    ) -> Tuple[int, int, int, int]:
+        text_w, text_h = cv2.getTextSize(
+            text=text,
+            fontFace=font,
+            fontScale=text_scale,
+            thickness=text_thickness,
+        )[0]
+        center_x, center_y = center_coordinates
+        return (
+            center_x - text_w // 2,
+            center_y - text_h // 2,
+            center_x + text_w // 2,
+            center_y + text_h // 2,
+        )
+
+    @staticmethod
+    def preprocess_and_validate_labels(
+        labels: Optional[List[str]], points_count: int, skeletons_count: int
+    ) -> np.ndarray:
+        if labels and len(labels) != points_count:
+            raise ValueError(
+                f"Number of labels ({len(labels)}) must match number of key points "
+                f"({points_count})."
+            )
+        if labels is None:
+            labels = [str(i) for i in range(points_count)]
+
+        return np.array(labels * skeletons_count)
+
+    @staticmethod
+    def preprocess_and_validate_colors(
+        colors: Optional[Union[Color, List[Color]]],
+        points_count: int,
+        skeletons_count: int,
+    ) -> np.ndarray:
+        if isinstance(colors, list) and len(colors) != points_count:
+            raise ValueError(
+                f"Number of colors ({len(colors)}) must match number of key points "
+                f"({points_count})."
+            )
+        return (
+            np.array(colors * skeletons_count)
+            if isinstance(colors, list)
+            else np.array([colors] * points_count * skeletons_count)
+        )

supervision/keypoint/core.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+from contextlib import suppress
 from dataclasses import dataclass, field
 from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
 
@@ -17,12 +18,13 @@ class KeyPoints:
     The `sv.KeyPoints` class in the Supervision library standardizes results from
     various keypoint detection and pose estimation models into a consistent format. This
     class simplifies data manipulation and filtering, providing a uniform API for
-    integration with Supervision annotators.
+    integration with Supervision [keypoints annotators](/keypoint/annotators).
 
     === "Ultralytics"
 
         Use [`sv.KeyPoints.from_ultralytics`](/keypoint/core/#supervision.keypoint.core.KeyPoints.from_ultralytics)
-        method, which accepts model results.
+        method, which accepts [YOLOv8](https://github.com/ultralytics/ultralytics)
+        pose result.
 
         ```python
         import cv2
@@ -31,20 +33,72 @@ class KeyPoints:
 
         image = cv2.imread(<SOURCE_IMAGE_PATH>)
         model = YOLO('yolov8s-pose.pt')
+
         result = model(image)[0]
         key_points = sv.KeyPoints.from_ultralytics(result)
         ```
 
+    === "Inference"
+
+        Use [`sv.KeyPoints.from_inference`](/keypoint/core/#supervision.keypoint.core.KeyPoints.from_inference)
+        method, which accepts [Inference](https://inference.roboflow.com/) pose result.
+
+        ```python
+        import cv2
+        import supervision as sv
+        from inference import get_model
+
+        image = cv2.imread(<SOURCE_IMAGE_PATH>)
+        model = get_model(model_id=<POSE_MODEL_ID>, api_key=<ROBOFLOW_API_KEY>)
+
+        result = model.infer(image)[0]
+        key_points = sv.KeyPoints.from_inference(result)
+        ```
+
+    === "MediaPipe"
+
+        Use [`sv.KeyPoints.from_mediapipe`](/keypoint/core/#supervision.keypoint.core.KeyPoints.from_mediapipe)
+        method, which accepts [MediaPipe](https://github.com/google-ai-edge/mediapipe)
+        pose result.
+
+        ```python
+        import cv2
+        import mediapipe as mp
+        import supervision as sv
+
+        image = cv2.imread(<SOURCE_IMAGE_PATH>)
+        image_height, image_width, _ = image.shape
+        mediapipe_image = mp.Image(
+            image_format=mp.ImageFormat.SRGB,
+            data=cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
+
+        options = mp.tasks.vision.PoseLandmarkerOptions(
+            base_options=mp.tasks.BaseOptions(
+                model_asset_path="pose_landmarker_heavy.task"
+            ),
+            running_mode=mp.tasks.vision.RunningMode.IMAGE,
+            num_poses=2)
+
+        PoseLandmarker = mp.tasks.vision.PoseLandmarker
+        with PoseLandmarker.create_from_options(options) as landmarker:
+            pose_landmarker_result = landmarker.detect(mediapipe_image)
+
+        key_points = sv.KeyPoints.from_mediapipe(
+            pose_landmarker_result, (image_width, image_height))
+        ```
+
     Attributes:
-        xy (np.ndarray): An array of shape `(n, 2)` containing
-            the bounding boxes coordinates in format `[x1, y1]`
+        xy (np.ndarray): An array of shape `(n, m, 2)` containing
+            `n` detected objects, each composed of `m` equally-sized
+            sets of keypoints, where each point is `[x, y]`.
         confidence (Optional[np.ndarray]): An array of shape
-            `(n,)` containing the confidence scores of the keypoint keypoints.
+            `(n, m)` containing the confidence scores of each keypoint.
         class_id (Optional[np.ndarray]): An array of shape
-            `(n,)` containing the class ids of the keypoint keypoints.
+            `(n,)` containing the class ids of the detected objects.
         data (Dict[str, Union[np.ndarray, List]]): A dictionary containing additional
             data where each key is a string representing the data type, and the value
-            is either a NumPy array or a list of corresponding data.
+            is either a NumPy array or a list of corresponding data of length `n`
+            (one entry per detected object).
     """  # noqa: E501 // docs
 
     xy: npt.NDArray[np.float32]
@@ -62,7 +116,7 @@ class KeyPoints:
 
     def __len__(self) -> int:
         """
-        Returns the number of keypoints in the keypoints object.
+        Returns the number of keypoints in the `sv.KeyPoints` object.
         """
         return len(self.xy)
 
@@ -80,7 +134,7 @@ class KeyPoints:
     ]:
         """
         Iterates over the Keypoint object and yield a tuple of
-        `(xy, confidence, class_id, data)` for each keypoint detection.
+        `(xy, confidence, class_id, data)` for each object detection.
         """
         for i in range(len(self.xy)):
             yield (
@@ -100,20 +154,228 @@ class KeyPoints:
             ]
         )
 
+    @classmethod
+    def from_inference(cls, inference_result: Union[dict, Any]) -> KeyPoints:
+        """
+        Create a `sv.KeyPoints` object from the [Roboflow](https://roboflow.com/)
+        API inference result or the [Inference](https://inference.roboflow.com/)
+        package results.
+
+        Args:
+            inference_result (dict, any): The result from the
+                Roboflow API or Inference package containing predictions with keypoints.
+
+        Returns:
+            A `sv.KeyPoints` object containing the keypoint coordinates, class IDs,
+                and class names, and confidences of each keypoint.
+
+        Examples:
+            ```python
+            import cv2
+            import supervision as sv
+            from inference import get_model
+
+            image = cv2.imread(<SOURCE_IMAGE_PATH>)
+            model = get_model(model_id=<POSE_MODEL_ID>, api_key=<ROBOFLOW_API_KEY>)
+
+            result = model.infer(image)[0]
+            key_points = sv.KeyPoints.from_inference(result)
+            ```
+
+            ```python
+            import cv2
+            import supervision as sv
+            from inference_sdk import InferenceHTTPClient
+
+            image = cv2.imread(<SOURCE_IMAGE_PATH>)
+            client = InferenceHTTPClient(
+                api_url="https://detect.roboflow.com",
+                api_key=<ROBOFLOW_API_KEY>
+            )
+
+            result = client.infer(image, model_id=<POSE_MODEL_ID>)
+            key_points = sv.KeyPoints.from_inference(result)
+            ```
+        """
+        if isinstance(inference_result, list):
+            raise ValueError(
+                "from_inference() operates on a single result at a time."
+                "You can retrieve it like so:  inference_result = model.infer(image)[0]"
+            )
+
+        with suppress(AttributeError):
+            inference_result = inference_result.dict(exclude_none=True, by_alias=True)
+
+        if not inference_result.get("predictions"):
+            return cls.empty()
+
+        xy = []
+        confidence = []
+        class_id = []
+        class_names = []
+
+        for prediction in inference_result["predictions"]:
+            prediction_xy = []
+            prediction_confidence = []
+            for keypoint in prediction["keypoints"]:
+                prediction_xy.append([keypoint["x"], keypoint["y"]])
+                prediction_confidence.append(keypoint["confidence"])
+            xy.append(prediction_xy)
+            confidence.append(prediction_confidence)
+
+            class_id.append(prediction["class_id"])
+            class_names.append(prediction["class"])
+
+        data = {CLASS_NAME_DATA_FIELD: np.array(class_names)}
+
+        return cls(
+            xy=np.array(xy, dtype=np.float32),
+            confidence=np.array(confidence, dtype=np.float32),
+            class_id=np.array(class_id, dtype=int),
+            data=data,
+        )
+
+    @classmethod
+    def from_mediapipe(
+        cls, mediapipe_results, resolution_wh: Tuple[int, int]
+    ) -> KeyPoints:
+        """
+        Creates a `sv.KeyPoints` instance from a
+        [MediaPipe](https://github.com/google-ai-edge/mediapipe)
+        pose landmark detection inference result.
+
+        Args:
+            mediapipe_results (Union[PoseLandmarkerResult, FaceLandmarkerResult, SolutionOutputs]):
+                The output results from Mediapipe. It support pose and face landmarks
+                from `PoseLandmaker`, `FaceLandmarker` and the legacy ones
+                from `Pose` and `FaceMesh`.
+            resolution_wh (Tuple[int, int]): A tuple of the form `(width, height)`
+                representing the resolution of the frame.
+
+        Returns:
+            A `sv.KeyPoints` object containing the keypoint coordinates and
+                confidences of each keypoint.
+
+        !!! tip
+            Before you start, download model bundles from the
+            [MediaPipe website](https://ai.google.dev/edge/mediapipe/solutions/vision/pose_landmarker/index#models).
+
+        Examples:
+            ```python
+            import cv2
+            import mediapipe as mp
+            import supervision as sv
+
+            image = cv2.imread(<SOURCE_IMAGE_PATH>)
+            image_height, image_width, _ = image.shape
+            mediapipe_image = mp.Image(
+                image_format=mp.ImageFormat.SRGB,
+                data=cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
+
+            options = mp.tasks.vision.PoseLandmarkerOptions(
+                base_options=mp.tasks.BaseOptions(
+                    model_asset_path="pose_landmarker_heavy.task"
+                ),
+                running_mode=mp.tasks.vision.RunningMode.IMAGE,
+                num_poses=2)
+
+            PoseLandmarker = mp.tasks.vision.PoseLandmarker
+            with PoseLandmarker.create_from_options(options) as landmarker:
+                pose_landmarker_result = landmarker.detect(mediapipe_image)
+
+            key_points = sv.KeyPoints.from_mediapipe(
+                pose_landmarker_result, (image_width, image_height))
+            ```
+
+            ```python
+            import cv2
+            import mediapipe as mp
+            import supervision as sv
+
+            image = cv2.imread(<SOURCE_IMAGE_PATH>)
+            image_height, image_width, _ = image.shape
+            mediapipe_image = mp.Image(
+                image_format=mp.ImageFormat.SRGB,
+                data=cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
+
+            options = mp.tasks.vision.FaceLandmarkerOptions(
+                base_options=mp.tasks.BaseOptions(
+                    model_asset_path="face_landmarker.task"
+                ),
+                output_face_blendshapes=True,
+                output_facial_transformation_matrixes=True,
+                num_faces=2)
+
+            FaceLandmarker = mp.tasks.vision.FaceLandmarker
+            with FaceLandmarker.create_from_options(options) as landmarker:
+                face_landmarker_result = landmarker.detect(mediapipe_image)
+
+            key_points = sv.KeyPoints.from_mediapipe(
+                face_landmarker_result, (image_width, image_height))
+            ```
+        """  # noqa: E501 // docs
+        if hasattr(mediapipe_results, "pose_landmarks"):
+            results = mediapipe_results.pose_landmarks
+            if not isinstance(mediapipe_results.pose_landmarks, list):
+                if mediapipe_results.pose_landmarks is None:
+                    results = []
+                else:
+                    results = [
+                        [
+                            landmark
+                            for landmark in mediapipe_results.pose_landmarks.landmark
+                        ]
+                    ]
+        elif hasattr(mediapipe_results, "face_landmarks"):
+            results = mediapipe_results.face_landmarks
+        elif hasattr(mediapipe_results, "multi_face_landmarks"):
+            if mediapipe_results.multi_face_landmarks is None:
+                results = []
+            else:
+                results = [
+                    face_landmark.landmark
+                    for face_landmark in mediapipe_results.multi_face_landmarks
+                ]
+
+        if len(results) == 0:
+            return cls.empty()
+
+        xy = []
+        confidence = []
+        for pose in results:
+            prediction_xy = []
+            prediction_confidence = []
+            for landmark in pose:
+                keypoint_xy = [
+                    landmark.x * resolution_wh[0],
+                    landmark.y * resolution_wh[1],
+                ]
+                prediction_xy.append(keypoint_xy)
+                prediction_confidence.append(landmark.visibility)
+
+            xy.append(prediction_xy)
+            confidence.append(prediction_confidence)
+
+        return cls(
+            xy=np.array(xy, dtype=np.float32),
+            confidence=np.array(confidence, dtype=np.float32),
+        )
+
     @classmethod
     def from_ultralytics(cls, ultralytics_results) -> KeyPoints:
         """
-        Creates a Keypoints instance from a
-            [YOLOv8](https://github.com/ultralytics/ultralytics) inference result.
+        Creates a `sv.KeyPoints` instance from a
+        [YOLOv8](https://github.com/ultralytics/ultralytics) pose inference result.
 
         Args:
             ultralytics_results (ultralytics.engine.results.Keypoints):
                 The output Results instance from YOLOv8
 
         Returns:
-            KeyPoints: A new Keypoints object.
+            A `sv.KeyPoints` object containing the keypoint coordinates, class IDs,
+                and class names, and confidences of each keypoint.
 
-        Example:
+        Examples:
             ```python
             import cv2
             import supervision as sv
@@ -121,8 +383,9 @@ class KeyPoints:
 
             image = cv2.imread(<SOURCE_IMAGE_PATH>)
             model = YOLO('yolov8s-pose.pt')
+
             result = model(image)[0]
-            keypoints = sv.KeyPoints.from_ultralytics(result)
+            key_points = sv.KeyPoints.from_ultralytics(result)
             ```
         """
         if ultralytics_results.keypoints.xy.numel() == 0:
@@ -136,38 +399,155 @@ class KeyPoints:
         data = {CLASS_NAME_DATA_FIELD: class_names}
         return cls(xy, class_id, confidence, data)
 
-    def __getitem__(
-        self, index: Union[int, slice, List[int], np.ndarray, str]
-    ) -> Union["KeyPoints", List, np.ndarray, None]:
+    @classmethod
+    def from_yolo_nas(cls, yolo_nas_results) -> KeyPoints:
         """
-        Get a subset of the KeyPoints object or access an item from its data field.
-
-        When provided with an integer, slice, list of integers, or a numpy array, this
-        method returns a new KeyPoints object that represents a subset of the original
-        keypoints. When provided with a string, it accesses the corresponding item in
-        the data dictionary.
+        Create a `sv.KeyPoints` instance from a [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS-POSE.md)
+        pose inference results.
 
         Args:
-            index (Union[int, slice, List[int], np.ndarray, str]): The index, indices,
-                or key to access a subset of the KeyPoints or an item from the data.
+            yolo_nas_results (ImagePoseEstimationPrediction): The output object from
+                YOLO NAS.
 
         Returns:
-            Union[KeyPoints, Any]: A subset of the KeyPoints object or an item from
-                the data field.
+            A `sv.KeyPoints` object containing the keypoint coordinates, class IDs,
+                and class names, and confidences of each keypoint.
+
+        Examples:
+            ```python
+            import cv2
+            import torch
+            import supervision as sv
+            import super_gradients
+
+            image = cv2.imread(<SOURCE_IMAGE_PATH>)
+
+            device = "cuda" if torch.cuda.is_available() else "cpu"
+            model = super_gradients.training.models.get(
+                "yolo_nas_pose_s", pretrained_weights="coco_pose").to(device)
+
+            results = model.predict(image, conf=0.1)
+            key_points = sv.KeyPoints.from_yolo_nas(results)
+            ```
+        """  # noqa: E501 // docs
+        if len(yolo_nas_results.prediction.poses) == 0:
+            return cls.empty()
+
+        xy = yolo_nas_results.prediction.poses[:, :, :2]
+        confidence = yolo_nas_results.prediction.poses[:, :, 2]
+
+        # yolo_nas_results treats params differently.
+        # prediction.labels may not exist, whereas class_names might be None
+        if hasattr(yolo_nas_results.prediction, "labels"):
+            class_id = yolo_nas_results.prediction.labels  # np.array[int]
+        else:
+            class_id = None
+
+        data = {}
+        if class_id is not None and yolo_nas_results.class_names is not None:
+            class_names = []
+            for c_id in class_id:
+                name = yolo_nas_results.class_names[c_id]  # tuple[str]
+                class_names.append(name)
+            data[CLASS_NAME_DATA_FIELD] = class_names
+
+        return cls(
+            xy=xy,
+            confidence=confidence,
+            class_id=class_id,
+            data=data,
+        )
+
+    @classmethod
+    def from_detectron2(cls, detectron2_results) -> KeyPoints:
+        """
+        Create a `sv.KeyPoints` object from the
+        [Detectron2](https://github.com/facebookresearch/detectron2) inference result.
+
+        Args:
+            detectron2_results: The output of a
+                Detectron2 model containing instances with prediction data.
+
+        Returns:
+            A `sv.KeyPoints` object containing the keypoint coordinates, class IDs,
+                and class names, and confidences of each keypoint.
 
         Example:
+            ```python
+            import cv2
+            import supervision as sv
+            from detectron2.engine import DefaultPredictor
+            from detectron2.config import get_cfg
+
+
+            image = cv2.imread(<SOURCE_IMAGE_PATH>)
+            cfg = get_cfg()
+            cfg.merge_from_file(<CONFIG_PATH>)
+            cfg.MODEL.WEIGHTS = <WEIGHTS_PATH>
+            predictor = DefaultPredictor(cfg)
+
+            result = predictor(image)
+            keypoints = sv.KeyPoints.from_detectron2(result)
+            ```
+        """
+
+        if hasattr(detectron2_results["instances"], "pred_keypoints"):
+            if detectron2_results["instances"].pred_keypoints.cpu().numpy().size == 0:
+                return cls.empty()
+            return cls(
+                xy=detectron2_results["instances"]
+                .pred_keypoints.cpu()
+                .numpy()[:, :, :2],
+                confidence=detectron2_results["instances"]
+                .pred_keypoints.cpu()
+                .numpy()[:, :, 2],
+                class_id=detectron2_results["instances"]
+                .pred_classes.cpu()
+                .numpy()
+                .astype(int),
+            )
+        else:
+            return cls.empty()
+
+    def __getitem__(
+        self, index: Union[int, slice, List[int], np.ndarray, str]
+    ) -> Union[KeyPoints, List, np.ndarray, None]:
+        """
+        Get a subset of the `sv.KeyPoints` object or access an item from its data field.
+
+        When provided with an integer, slice, list of integers, or a numpy array, this
+        method returns a new `sv.KeyPoints` object that represents a subset of the
+        original `sv.KeyPoints`. When provided with a string, it accesses the
+        corresponding item in the data dictionary.
+
+        Args:
+            index (Union[int, slice, List[int], np.ndarray, str]): The index, indices,
+                or key to access a subset of the `sv.KeyPoints` or an item from the
+                data.
+
+        Returns:
+            A subset of the `sv.KeyPoints` object or an item from the data field.
+
+        Examples:
             ```python
             import supervision as sv
 
-            keypoints = sv.KeyPoints()
+            key_points = sv.KeyPoints()
 
-            first_detection = keypoints[0]
-            first_10_keypoints = keypoints[0:10]
-            some_keypoints = keypoints[[0, 2, 4]]
-            class_0_keypoints = keypoints[keypoints.class_id == 0]
-            high_confidence_keypoints = keypoints[keypoints.confidence > 0.5]
+            # access the first keypoint using an integer index
+            key_points[0]
 
-            feature_vector = keypoints['feature_vector']
+            # access the first 10 keypoints using index slice
+            key_points[0:10]
+
+            # access selected keypoints using a list of indices
+            key_points[[0, 2, 4]]
+
+            # access keypoints with selected class_id
+            key_points[key_points.class_id == 0]
+
+            # access keypoints with confidence greater than 0.5
+            key_points[key_points.confidence > 0.5]
             ```
         """
         if isinstance(index, str):
@@ -183,13 +563,13 @@ class KeyPoints:
 
     def __setitem__(self, key: str, value: Union[np.ndarray, List]):
         """
-        Set a value in the data dictionary of the KeyPoints object.
+        Set a value in the data dictionary of the `sv.KeyPoints` object.
 
         Args:
             key (str): The key in the data dictionary to set.
             value (Union[np.ndarray, List]): The value to set for the key.
 
-        Example:
+        Examples:
             ```python
             import cv2
             import supervision as sv
@@ -201,7 +581,7 @@ class KeyPoints:
             result = model(image)[0]
             keypoints = sv.KeyPoints.from_ultralytics(result)
 
-            keypoints['names'] = [
+            keypoints['class_name'] = [
                  model.model.names[class_id]
                  for class_id
                  in keypoints.class_id
@@ -222,12 +602,13 @@ class KeyPoints:
         Create an empty Keypoints object with no keypoints.
 
         Returns:
-            (KeyPoints): An empty Keypoints object.
+            An empty `sv.KeyPoints` object.
 
-        Example:
+        Examples:
             ```python
-            from supervision import Keypoints
-            empty_keypoints = Keypoints.empty()
+            import supervision as sv
+
+            key_points = sv.KeyPoints.empty()
             ```
         """
         return cls(xy=np.empty((0, 0, 2), dtype=np.float32))

supervision/metrics/__init__.py

@@ -0,0 +1,10 @@
+from supervision.metrics.core import CLASS_ID_NONE, Metric, MetricTarget
+from supervision.metrics.mean_average_precision import (
+    MeanAveragePrecision,
+    MeanAveragePrecisionResult,
+)
+from supervision.metrics.utils.object_size import (
+    ObjectSizeCategory,
+    get_detection_size_category,
+    get_object_size_category,
+)

supervision/metrics/core.py

@@ -0,0 +1,50 @@
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from enum import Enum
+from typing import Any
+
+CLASS_ID_NONE = -1
+"""Used by metrics module as class ID, when none is present"""
+
+
+class Metric(ABC):
+    """
+    The base class for all supervision metrics.
+    """
+
+    @abstractmethod
+    def update(self, *args, **kwargs) -> "Metric":
+        """
+        Add data to the metric, without computing the result.
+        Return the metric itself to allow method chaining.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def reset(self) -> None:
+        """
+        Reset internal metric state.
+        """
+        raise NotImplementedError
+
+    @abstractmethod
+    def compute(self, *args, **kwargs) -> Any:
+        """
+        Compute the metric from the internal state and return the result.
+        """
+        raise NotImplementedError
+
+
+class MetricTarget(Enum):
+    """
+    Specifies what type of detection is used to compute the metric.
+
+    * BOXES: xyxy bounding boxes
+    * MASKS: Binary masks
+    * ORIENTED_BOUNDING_BOXES: Oriented bounding boxes (OBB)
+    """
+
+    BOXES = "boxes"
+    MASKS = "masks"
+    ORIENTED_BOUNDING_BOXES = "obb"

supervision/metrics/detection.py

@@ -391,11 +391,10 @@ class ConfusionMatrix:
             ```
         """
         predictions, targets = [], []
-        for img_name, img in dataset.images.items():
-            predictions_batch = callback(img)
+        for _, image, annotation in dataset:
+            predictions_batch = callback(image)
             predictions.append(predictions_batch)
-            targets_batch = dataset.annotations[img_name]
-            targets.append(targets_batch)
+            targets.append(annotation)
         return cls.from_detections(
             predictions=predictions,
             targets=targets,
@@ -604,11 +603,10 @@ class MeanAveragePrecision:
             ```
         """
         predictions, targets = [], []
-        for img_name, img in dataset.images.items():
-            predictions_batch = callback(img)
+        for _, image, annotation in dataset:
+            predictions_batch = callback(image)
             predictions.append(predictions_batch)
-            targets_batch = dataset.annotations[img_name]
-            targets.append(targets_batch)
+            targets.append(annotation)
         return cls.from_detections(
             predictions=predictions,
             targets=targets,
@@ -666,12 +664,12 @@ class MeanAveragePrecision:
                 np.array([[1.0, 1.0, 2.0, 2.0, 2, 0.8]])
             ]
 
-            mean_average_precison = sv.MeanAveragePrecision.from_tensors(
+            mean_average_precision = sv.MeanAveragePrecision.from_tensors(
                 predictions=predictions,
                 targets=targets,
             )
 
-            print(mean_average_precison.map50_95)
+            print(mean_average_precision.map50_95)
             # 0.6649
             ```
         """
@@ -808,7 +806,7 @@ class MeanAveragePrecision:
             prediction_confidence (np.ndarray): Objectness value from 0-1.
             prediction_class_ids (np.ndarray): Predicted object classes.
             true_class_ids (np.ndarray): True object classes.
-            eps (float, optional): Small value to prevent division by zero.
+            eps (float): Small value to prevent division by zero.
 
         Returns:
             np.ndarray: Average precision for different IoU levels.

supervision/metrics/mean_average_precision.py

@@ -0,0 +1,552 @@
+from __future__ import annotations
+
+from copy import deepcopy
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, List, Optional, Union
+
+import numpy as np
+from matplotlib import pyplot as plt
+
+from supervision.config import ORIENTED_BOX_COORDINATES
+from supervision.detection.core import Detections
+from supervision.detection.utils import box_iou_batch, mask_iou_batch
+from supervision.draw.color import LEGACY_COLOR_PALETTE
+from supervision.metrics.core import Metric, MetricTarget
+from supervision.metrics.utils.object_size import (
+    ObjectSizeCategory,
+    get_detection_size_category,
+)
+from supervision.metrics.utils.utils import ensure_pandas_installed
+
+if TYPE_CHECKING:
+    import pandas as pd
+
+
+class MeanAveragePrecision(Metric):
+    def __init__(
+        self,
+        metric_target: MetricTarget = MetricTarget.BOXES,
+        class_agnostic: bool = False,
+    ):
+        """
+        Initialize the Mean Average Precision metric.
+
+        Args:
+            metric_target (MetricTarget): The type of detection data to use.
+            class_agnostic (bool): Whether to treat all data as a single class.
+        """
+        self._metric_target = metric_target
+        if self._metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+            raise NotImplementedError(
+                "Mean Average Precision is not implemented for oriented bounding boxes."
+            )
+
+        self._class_agnostic = class_agnostic
+
+        self._predictions_list: List[Detections] = []
+        self._targets_list: List[Detections] = []
+
+    def reset(self) -> None:
+        self._predictions_list = []
+        self._targets_list = []
+
+    def update(
+        self,
+        predictions: Union[Detections, List[Detections]],
+        targets: Union[Detections, List[Detections]],
+    ) -> MeanAveragePrecision:
+        """
+        Add new predictions and targets to the metric, but do not compute the result.
+
+        Args:
+            predictions (Union[Detections, List[Detections]]): The predicted detections.
+            targets (Union[Detections, List[Detections]]): The ground-truth detections.
+
+        Returns:
+            (MeanAveragePrecision): The updated metric instance.
+        """
+        if not isinstance(predictions, list):
+            predictions = [predictions]
+        if not isinstance(targets, list):
+            targets = [targets]
+
+        if len(predictions) != len(targets):
+            raise ValueError(
+                f"The number of predictions ({len(predictions)}) and"
+                f" targets ({len(targets)}) during the update must be the same."
+            )
+
+        self._predictions_list.extend(predictions)
+        self._targets_list.extend(targets)
+
+        return self
+
+    def compute(
+        self,
+    ) -> MeanAveragePrecisionResult:
+        """
+        Calculate Mean Average Precision based on predicted and ground-truth
+            detections at different thresholds.
+
+        Returns:
+            (MeanAveragePrecisionResult): New instance of MeanAveragePrecision.
+
+        Example:
+            ```python
+            import supervision as sv
+            from supervision.metrics import MeanAveragePrecision
+
+            predictions = sv.Detections(...)
+            targets = sv.Detections(...)
+
+            map_metric = MeanAveragePrecision()
+            map_result = map_metric.update(predictions, targets).compute()
+
+            print(map_result)
+            print(map_result.map50_95)
+            map_result.plot()
+            ```
+        """
+        result = self._compute(self._predictions_list, self._targets_list)
+
+        small_predictions = []
+        small_targets = []
+        for predictions, targets in zip(self._predictions_list, self._targets_list):
+            small_predictions.append(
+                self._filter_detections_by_size(predictions, ObjectSizeCategory.SMALL)
+            )
+            small_targets.append(
+                self._filter_detections_by_size(targets, ObjectSizeCategory.SMALL)
+            )
+        result.small_objects = self._compute(small_predictions, small_targets)
+
+        medium_predictions = []
+        medium_targets = []
+        for predictions, targets in zip(self._predictions_list, self._targets_list):
+            medium_predictions.append(
+                self._filter_detections_by_size(predictions, ObjectSizeCategory.MEDIUM)
+            )
+            medium_targets.append(
+                self._filter_detections_by_size(targets, ObjectSizeCategory.MEDIUM)
+            )
+        result.medium_objects = self._compute(medium_predictions, medium_targets)
+
+        large_predictions = []
+        large_targets = []
+        for predictions, targets in zip(self._predictions_list, self._targets_list):
+            large_predictions.append(
+                self._filter_detections_by_size(predictions, ObjectSizeCategory.LARGE)
+            )
+            large_targets.append(
+                self._filter_detections_by_size(targets, ObjectSizeCategory.LARGE)
+            )
+        result.large_objects = self._compute(large_predictions, large_targets)
+
+        return result
+
+    def _compute(
+        self,
+        predictions_list: List[Detections],
+        targets_list: List[Detections],
+    ) -> MeanAveragePrecisionResult:
+        iou_thresholds = np.linspace(0.5, 0.95, 10)
+        stats = []
+
+        for predictions, targets in zip(predictions_list, targets_list):
+            prediction_contents = self._detections_content(predictions)
+            target_contents = self._detections_content(targets)
+
+            if len(targets) > 0:
+                if len(predictions) == 0:
+                    stats.append(
+                        (
+                            np.zeros((0, iou_thresholds.size), dtype=bool),
+                            np.zeros((0,), dtype=np.float32),
+                            np.zeros((0,), dtype=int),
+                            targets.class_id,
+                        )
+                    )
+
+                else:
+                    if self._metric_target == MetricTarget.BOXES:
+                        iou = box_iou_batch(target_contents, prediction_contents)
+                    elif self._metric_target == MetricTarget.MASKS:
+                        iou = mask_iou_batch(target_contents, prediction_contents)
+                    else:
+                        raise NotImplementedError(
+                            "Unsupported metric target for IoU calculation"
+                        )
+
+                    matches = self._match_detection_batch(
+                        predictions.class_id, targets.class_id, iou, iou_thresholds
+                    )
+                    stats.append(
+                        (
+                            matches,
+                            predictions.confidence,
+                            predictions.class_id,
+                            targets.class_id,
+                        )
+                    )
+
+        # Compute average precisions if any matches exist
+        if stats:
+            concatenated_stats = [np.concatenate(items, 0) for items in zip(*stats)]
+            average_precisions = self._average_precisions_per_class(*concatenated_stats)
+            map50 = average_precisions[:, 0].mean()
+            map75 = average_precisions[:, 5].mean()
+            map50_95 = average_precisions.mean()
+        else:
+            map50, map75, map50_95 = 0, 0, 0
+            average_precisions = np.empty((0, len(iou_thresholds)), dtype=np.float32)
+
+        return MeanAveragePrecisionResult(
+            iou_thresholds=iou_thresholds,
+            map50_95=map50_95,
+            map50=map50,
+            map75=map75,
+            per_class_ap50_95=average_precisions,
+            metric_target=self._metric_target,
+        )
+
+    @staticmethod
+    def _compute_average_precision(recall: np.ndarray, precision: np.ndarray) -> float:
+        """
+        Compute the average precision using 101-point interpolation (COCO), given
+            the recall and precision curves.
+
+        Args:
+            recall (np.ndarray): The recall curve.
+            precision (np.ndarray): The precision curve.
+
+        Returns:
+            (float): Average precision.
+        """
+        if len(recall) == 0 and len(precision) == 0:
+            return 0.0
+
+        recall_levels = np.linspace(0, 1, 101)
+        precision_levels = np.zeros_like(recall_levels)
+        for r, p in zip(recall[::-1], precision[::-1]):
+            precision_levels[recall_levels <= r] = p
+
+        average_precision = (1 / 100 * precision_levels).sum()
+        return average_precision
+
+    @staticmethod
+    def _match_detection_batch(
+        predictions_classes: np.ndarray,
+        target_classes: np.ndarray,
+        iou: np.ndarray,
+        iou_thresholds: np.ndarray,
+    ) -> np.ndarray:
+        num_predictions, num_iou_levels = (
+            predictions_classes.shape[0],
+            iou_thresholds.shape[0],
+        )
+        correct = np.zeros((num_predictions, num_iou_levels), dtype=bool)
+        correct_class = target_classes[:, None] == predictions_classes
+
+        for i, iou_level in enumerate(iou_thresholds):
+            matched_indices = np.where((iou >= iou_level) & correct_class)
+
+            if matched_indices[0].shape[0]:
+                combined_indices = np.stack(matched_indices, axis=1)
+                iou_values = iou[matched_indices][:, None]
+                matches = np.hstack([combined_indices, iou_values])
+
+                if matched_indices[0].shape[0] > 1:
+                    matches = matches[matches[:, 2].argsort()[::-1]]
+                    matches = matches[np.unique(matches[:, 1], return_index=True)[1]]
+                    matches = matches[np.unique(matches[:, 0], return_index=True)[1]]
+
+                correct[matches[:, 1].astype(int), i] = True
+
+        return correct
+
+    @staticmethod
+    def _average_precisions_per_class(
+        matches: np.ndarray,
+        prediction_confidence: np.ndarray,
+        prediction_class_ids: np.ndarray,
+        true_class_ids: np.ndarray,
+    ) -> np.ndarray:
+        """
+        Compute the average precision, given the recall and precision curves.
+        Source: https://github.com/rafaelpadilla/Object-Detection-Metrics.
+
+        Args:
+            matches (np.ndarray): True positives.
+            prediction_confidence (np.ndarray): Objectness value from 0-1.
+            prediction_class_ids (np.ndarray): Predicted object classes.
+            true_class_ids (np.ndarray): True object classes.
+            eps (float, optional): Small value to prevent division by zero.
+
+        Returns:
+            (np.ndarray): Average precision for different IoU levels.
+        """
+        eps = 1e-16
+
+        sorted_indices = np.argsort(-prediction_confidence)
+        matches = matches[sorted_indices]
+        prediction_class_ids = prediction_class_ids[sorted_indices]
+
+        unique_classes, class_counts = np.unique(true_class_ids, return_counts=True)
+        num_classes = unique_classes.shape[0]
+
+        average_precisions = np.zeros((num_classes, matches.shape[1]))
+
+        for class_idx, class_id in enumerate(unique_classes):
+            is_class = prediction_class_ids == class_id
+            total_true = class_counts[class_idx]
+            total_prediction = is_class.sum()
+
+            if total_prediction == 0 or total_true == 0:
+                continue
+
+            false_positives = (1 - matches[is_class]).cumsum(0)
+            true_positives = matches[is_class].cumsum(0)
+            false_negatives = total_true - true_positives
+
+            recall = true_positives / (true_positives + false_negatives + eps)
+            precision = true_positives / (true_positives + false_positives)
+
+            for iou_level_idx in range(matches.shape[1]):
+                average_precisions[class_idx, iou_level_idx] = (
+                    MeanAveragePrecision._compute_average_precision(
+                        recall[:, iou_level_idx], precision[:, iou_level_idx]
+                    )
+                )
+
+        return average_precisions
+
+    def _detections_content(self, detections: Detections) -> np.ndarray:
+        """Return boxes, masks or oriented bounding boxes from detections."""
+        if self._metric_target == MetricTarget.BOXES:
+            return detections.xyxy
+        if self._metric_target == MetricTarget.MASKS:
+            return (
+                detections.mask
+                if detections.mask is not None
+                else self._make_empty_content()
+            )
+        if self._metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+            if obb := detections.data.get(ORIENTED_BOX_COORDINATES):
+                return np.ndarray(obb, dtype=np.float32)
+            return self._make_empty_content()
+        raise ValueError(f"Invalid metric target: {self._metric_target}")
+
+    def _make_empty_content(self) -> np.ndarray:
+        if self._metric_target == MetricTarget.BOXES:
+            return np.empty((0, 4), dtype=np.float32)
+        if self._metric_target == MetricTarget.MASKS:
+            return np.empty((0, 0, 0), dtype=bool)
+        if self._metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+            return np.empty((0, 8), dtype=np.float32)
+        raise ValueError(f"Invalid metric target: {self._metric_target}")
+
+    def _filter_detections_by_size(
+        self, detections: Detections, size_category: ObjectSizeCategory
+    ) -> Detections:
+        """Return a copy of detections with contents filtered by object size."""
+        new_detections = deepcopy(detections)
+        if detections.is_empty() or size_category == ObjectSizeCategory.ANY:
+            return new_detections
+
+        sizes = get_detection_size_category(new_detections, self._metric_target)
+        size_mask = sizes == size_category.value
+
+        new_detections.xyxy = new_detections.xyxy[size_mask]
+        if new_detections.mask is not None:
+            new_detections.mask = new_detections.mask[size_mask]
+        if new_detections.class_id is not None:
+            new_detections.class_id = new_detections.class_id[size_mask]
+        if new_detections.confidence is not None:
+            new_detections.confidence = new_detections.confidence[size_mask]
+        if new_detections.tracker_id is not None:
+            new_detections.tracker_id = new_detections.tracker_id[size_mask]
+        if new_detections.data is not None:
+            for key, value in new_detections.data.items():
+                new_detections.data[key] = np.array(value)[size_mask]
+
+        return new_detections
+
+
+@dataclass
+class MeanAveragePrecisionResult:
+    iou_thresholds: np.ndarray
+    """Array of IoU thresholds used in the calculations"""
+    map50_95: float
+    """Mean Average Precision over IoU thresholds from 0.5 to 0.95"""
+
+    map50: float
+    """Mean Average Precision at IoU threshold of 0.5"""
+
+    map75: float
+    """Mean Average Precision at IoU threshold of 0.75"""
+
+    per_class_ap50_95: np.ndarray
+    """Average precision for each class at different IoU thresholds"""
+
+    metric_target: MetricTarget
+    """
+    Defines the type of data used for the metric - boxes, masks or
+    oriented bounding boxes.
+    """
+
+    small_objects: Optional[MeanAveragePrecisionResult] = None
+    """Mean Average Precision results for small objects"""
+
+    medium_objects: Optional[MeanAveragePrecisionResult] = None
+    """Mean Average Precision results for medium objects"""
+
+    large_objects: Optional[MeanAveragePrecisionResult] = None
+    """Mean Average Precision results for large objects"""
+
+    def __str__(self) -> str:
+        """
+        Format the mAP results as a pretty string.
+
+        Example:
+            ```python
+            print(map_result)
+            ```
+        """
+
+        out_str = (
+            f"{self.__class__.__name__}:\n"
+            f"iou_thresholds: {self.iou_thresholds}\n"
+            f"map50_95:  {self.map50_95:.4f}\n"
+            f"map50:     {self.map50:.4f}\n"
+            f"map75:     {self.map75:.4f}\n"
+            f"per_class_ap50_95:"
+        )
+
+        for class_id, ap in enumerate(self.per_class_ap50_95):
+            out_str += f"\n  {class_id}:  {ap}"
+
+        indent = "  "
+        if self.small_objects is not None:
+            indented_str = indent + str(self.small_objects).replace("\n", f"\n{indent}")
+            out_str += f"\nSmall objects:\n{indented_str}"
+        if self.medium_objects is not None:
+            indented_str = indent + str(self.medium_objects).replace(
+                "\n", f"\n{indent}"
+            )
+            out_str += f"\nMedium objects:\n{indented_str}"
+        if self.large_objects is not None:
+            indented_str = indent + str(self.large_objects).replace("\n", f"\n{indent}")
+            out_str += f"\nLarge objects:\n{indented_str}"
+
+        return out_str
+
+    def to_pandas(self) -> "pd.DataFrame":
+        """
+        Convert the result to a pandas DataFrame.
+
+        Returns:
+            (pd.DataFrame): The result as a DataFrame.
+        """
+        ensure_pandas_installed()
+        import pandas as pd
+
+        pandas_data = {
+            "mAP_50_95": self.map50_95,
+            "mAP_50": self.map50,
+            "mAP_75": self.map75,
+        }
+        if self.small_objects is not None:
+            small_objects_df = self.small_objects.to_pandas()
+            for key, value in small_objects_df.items():
+                pandas_data[f"small_objects_{key}"] = value
+        if self.medium_objects is not None:
+            medium_objects_df = self.medium_objects.to_pandas()
+            for key, value in medium_objects_df.items():
+                pandas_data[f"medium_objects_{key}"] = value
+        if self.large_objects is not None:
+            large_objects_df = self.large_objects.to_pandas()
+            for key, value in large_objects_df.items():
+                pandas_data[f"large_objects_{key}"] = value
+
+        # Average precisions are currently not included in the DataFrame.
+
+        return pd.DataFrame(
+            pandas_data,
+            index=[0],
+        )
+
+    def plot(self):
+        """
+        Plot the mAP results.
+        """
+
+        labels = ["mAP_50_95", "mAP_50", "mAP_75"]
+        values = [self.map50_95, self.map50, self.map75]
+        colors = [LEGACY_COLOR_PALETTE[0]] * 3
+
+        if self.small_objects is not None:
+            labels += [
+                "small_objects_mAP_50_95",
+                "small_objects_mAP_50",
+                "small_objects_mAP_75",
+            ]
+            values += [
+                self.small_objects.map50_95,
+                self.small_objects.map50,
+                self.small_objects.map75,
+            ]
+            colors += [LEGACY_COLOR_PALETTE[3]] * 3
+
+        if self.medium_objects is not None:
+            labels += [
+                "medium_objects_mAP_50_95",
+                "medium_objects_mAP_50",
+                "medium_objects_mAP_75",
+            ]
+            values += [
+                self.medium_objects.map50_95,
+                self.medium_objects.map50,
+                self.medium_objects.map75,
+            ]
+            colors += [LEGACY_COLOR_PALETTE[2]] * 3
+
+        if self.large_objects is not None:
+            labels += [
+                "large_objects_mAP_50_95",
+                "large_objects_mAP_50",
+                "large_objects_mAP_75",
+            ]
+            values += [
+                self.large_objects.map50_95,
+                self.large_objects.map50,
+                self.large_objects.map75,
+            ]
+            colors += [LEGACY_COLOR_PALETTE[4]] * 3
+
+        plt.rcParams["font.family"] = "monospace"
+
+        _, ax = plt.subplots(figsize=(10, 6))
+        ax.set_ylim(0, 1)
+        ax.set_ylabel("Value", fontweight="bold")
+        ax.set_title("Mean Average Precision", fontweight="bold")
+
+        x_positions = range(len(labels))
+        bars = ax.bar(x_positions, values, color=colors, align="center")
+
+        ax.set_xticks(x_positions)
+        ax.set_xticklabels(labels, rotation=45, ha="right")
+
+        for bar in bars:
+            y_value = bar.get_height()
+            ax.text(
+                bar.get_x() + bar.get_width() / 2,
+                y_value + 0.02,
+                f"{y_value:.2f}",
+                ha="center",
+                va="bottom",
+            )
+
+        plt.rcParams["font.family"] = "sans-serif"
+
+        plt.tight_layout()
+        plt.show()

supervision/metrics/utils/internal_data_store.py

@@ -0,0 +1,300 @@
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, List, Optional, Set, Tuple
+
+import numpy as np
+import numpy.typing as npt
+
+from supervision.config import ORIENTED_BOX_COORDINATES
+from supervision.metrics.core import CLASS_ID_NONE, MetricTarget
+from supervision.metrics.utils.object_size import (
+    ObjectSizeCategory,
+    get_object_size_category,
+)
+from supervision.metrics.utils.utils import unify_pad_masks_shape
+
+if TYPE_CHECKING:
+    from supervision.detection.core import Detections
+
+
+class MetricData:
+    """
+    A container for detection contents, decouple from `Detections`.
+    """
+
+    def __init__(self, metric_target: MetricTarget, class_agnostic: bool = False):
+        self._metric_target = metric_target
+        self._class_agnostic = class_agnostic
+        self._content_list: List[npt.NDArray] = []
+        self._confidence_list: List[npt.NDArray[np.float32]] = []
+        self._class_id_list: List[npt.NDArray[np.int_]] = []
+
+    def update(self, detections: Detections):
+        """Add new detections to the store."""
+        if detections.is_empty() or len(detections) == 0:
+            return
+
+        # Relies on Detections to ensure that member vars are equal in length
+        # or None/empty.
+        self._validate_new_entry(detections)
+
+        new_content = self._detections_content(detections)
+        self._validate_shape(new_content)
+        self._content_list.append(new_content)
+
+        if detections.confidence is not None:
+            self._confidence_list.append(detections.confidence)
+
+        if detections.class_id is not None:
+            if self._class_agnostic:
+                self._class_id_list.append(
+                    np.full(len(detections.class_id), CLASS_ID_NONE, dtype=int)
+                )
+            else:
+                self._class_id_list.append(detections.class_id)
+
+    def get_classes(self) -> Set[int]:
+        """Return all class IDs."""
+        if self._class_agnostic:
+            return {CLASS_ID_NONE}
+        return set.union(*[set(class_id) for class_id in self._class_id_list])
+
+    def get(
+        self, class_id: Optional[int] = None, size_category=ObjectSizeCategory.ANY
+    ) -> Tuple[npt.NDArray, npt.NDArray, npt.NDArray]:
+        """
+        Get the contents, class_ids and confidences, optionally filtered by
+        class_id and/or size.
+
+        Args:
+            class_id (Optional[int]): The class ID for the data to retrieve
+            size_category (ObjectSizeCategory): The size of the objects to retrieve.
+
+        Returns:
+            (np.ndarray): Boxes, masks or obb, all stacked in a single array.
+        """
+        if self._metric_target in [
+            MetricTarget.BOXES,
+            MetricTarget.ORIENTED_BOUNDING_BOXES,
+        ]:
+            self._merge_boxes()
+            if len(self._content_list) == 0:
+                content = self._make_empty_content()
+            else:
+                content = self._content_list[0]
+
+        elif self._metric_target == MetricTarget.MASKS:
+            self._merge_masks()
+            if len(self._content_list) == 0:
+                content = self._make_empty_content()
+            else:
+                content = self._content_list[0]
+
+        size_mask = np.full(len(content), True)
+        if size_category != ObjectSizeCategory.ANY:
+            sizes = get_object_size_category(content, self._metric_target)
+            size_mask = sizes == size_category.value
+
+        class_mask = np.full(len(content), True)
+        if class_id is not None:
+            self._merge_class_id()
+            class_ids = self._class_id_list[0]
+            class_mask = class_ids == class_id
+
+        content = content[(size_mask & class_mask)]
+
+        self._merge_class_id()
+        if len(self._class_id_list) == 0:
+            class_ids = np.array([], dtype=int)
+        else:
+            class_ids = self._class_id_list[0]
+        if len(class_ids) > 0:
+            class_ids = class_ids[(size_mask & class_mask)]
+
+        self._merge_confidence()
+        if len(self._confidence_list) == 0:
+            confidences = np.array([], dtype=float)
+        else:
+            confidences = self._confidence_list[0]
+        if len(confidences) > 0:
+            confidences = confidences[(size_mask & class_mask)]
+
+        return content, class_ids, confidences
+
+    def get_class_id(self) -> npt.NDArray[np.int_]:
+        self._merge_class_id()
+        if len(self._class_id_list) == 0:
+            return np.array([], dtype=int)
+        return self._class_id_list[0]
+
+    def get_confidence(self) -> npt.NDArray[np.float32]:
+        self._merge_confidence()
+        if len(self._confidence_list) == 0:
+            return np.array([], dtype=float)
+        return self._confidence_list[0]
+
+    def _detections_content(self, detections: Detections) -> npt.NDArray:
+        """Return boxes, masks or oriented bounding boxes from detections."""
+        if self._metric_target == MetricTarget.BOXES:
+            return detections.xyxy
+        if self._metric_target == MetricTarget.MASKS:
+            return (
+                detections.mask
+                if detections.mask is not None
+                else self._make_empty_content()
+            )
+        if self._metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+            if obb := detections.data.get(ORIENTED_BOX_COORDINATES):
+                return np.ndarray(obb, dtype=np.float32)
+            return self._make_empty_content()
+        raise ValueError(f"Invalid metric target: {self._metric_target}")
+
+    def _make_empty_content(self) -> npt.NDArray:
+        if self._metric_target == MetricTarget.BOXES:
+            return np.empty((0, 4), dtype=np.float32)
+        if self._metric_target == MetricTarget.MASKS:
+            return np.empty((0, 0, 0), dtype=bool)
+        if self._metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+            return np.empty((0, 8), dtype=np.float32)
+        raise ValueError(f"Invalid metric target: {self._metric_target}")
+
+    def _validate_new_entry(self, detections: Detections) -> None:
+        if (
+            len(self._content_list) == 0
+            or detections.is_empty()
+            or len(detections) == 0
+        ):
+            return
+
+        is_self_class_empty = len(self._class_id_list) == 0
+        is_self_confidence_empty = len(self._confidence_list) == 0
+        is_detection_class_empty = (
+            detections.class_id is None or len(detections.class_id) == 0
+        )
+        is_detection_confidence_empty = (
+            detections.confidence is None or len(detections.confidence) == 0
+        )
+
+        if is_self_confidence_empty and not is_detection_confidence_empty:
+            raise ValueError(
+                "Previously stored detections without confidence,"
+                " but new detections have it."
+            )
+        if is_self_class_empty and not is_detection_class_empty:
+            raise ValueError(
+                "Previously stored detections without class ID,"
+                " but new detections have it."
+            )
+        if not is_self_class_empty and is_detection_class_empty:
+            raise ValueError(
+                "Started storing detections with class ID, but"
+                " new detections do not have it."
+            )
+        if not is_self_confidence_empty and is_detection_confidence_empty:
+            raise ValueError(
+                "Started storing detections with confidence,"
+                " but new detections do not have it."
+            )
+
+    def _validate_shape(self, data: npt.NDArray) -> None:
+        if self._metric_target == MetricTarget.BOXES:
+            if len(data.shape) != 2 or data.shape[1] != 4:
+                raise ValueError(f"Invalid xyxy shape: {data.shape}. Expected: (N, 4)")
+        elif self._metric_target == MetricTarget.MASKS:
+            if len(data.shape) != 3:
+                raise ValueError(
+                    f"Invalid mask shape: {data.shape}. Expected: (N, H, W)"
+                )
+        elif self._metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+            if len(data.shape) != 2 or data.shape[1] != 8:
+                raise ValueError(f"Invalid obb shape: {data.shape}. Expected: (N, 8)")
+        else:
+            raise ValueError(f"Invalid metric target: {self._metric_target}")
+
+    def _merge_boxes(self):
+        """Merge all boxes into a single array."""
+        if self._metric_target not in (
+            MetricTarget.BOXES,
+            MetricTarget.ORIENTED_BOUNDING_BOXES,
+        ):
+            raise ValueError("Invalid metric target for merging boxes")
+        if len(self._content_list) < 2:
+            return
+        self._content_list = [np.vstack(self._content_list)]
+
+    def _merge_masks(self):
+        """Merge all masks into a single array."""
+        if self._metric_target != MetricTarget.MASKS:
+            raise ValueError("Invalid metric target for merging masks")
+        if len(self._content_list) < 2:
+            return
+
+        mew_mask_list = unify_pad_masks_shape(*self._content_list)
+        self._content_list = [np.vstack(mew_mask_list)]
+
+    def _merge_class_id(self):
+        """Merge all class IDs into a single array."""
+        if len(self._class_id_list) < 2:
+            return
+        self._class_id_list = [np.hstack(self._class_id_list)]
+
+    def _merge_confidence(self):
+        """Merge all confidences into a single array."""
+        if len(self._confidence_list) < 2:
+            return
+        self._confidence_list = [np.hstack(self._confidence_list)]
+
+
+class MetricDataStore:
+    """
+    Stores internal data for metrics.
+
+    Provides a class-agnostic way to access it.
+    """
+
+    def __init__(self, metric_target: MetricTarget, class_agnostic: bool = False):
+        self._metric_target = metric_target
+        self._class_agnostic = class_agnostic
+        self._data_1: MetricData
+        self._data_2: MetricData
+        self.reset()
+
+    def reset(self) -> None:
+        self._data_1 = MetricData(self._metric_target, self._class_agnostic)
+        self._data_2 = MetricData(self._metric_target, self._class_agnostic)
+
+    def update(self, data_1: Detections, data_2: Detections) -> None:
+        """
+        Add new data to the store.
+
+        Use sv.Detections.empty() if only one set of data is available.
+        """
+        self._data_1.update(data_1)
+        self._data_2.update(data_2)
+
+    def get(
+        self,
+        class_id: Optional[int] = None,
+        size_category: ObjectSizeCategory = ObjectSizeCategory.ANY,
+    ) -> Tuple[
+        Tuple[npt.NDArray, npt.NDArray, npt.NDArray],
+        Tuple[npt.NDArray, npt.NDArray, npt.NDArray],
+    ]:
+        """
+        Get the data for both sets, optionally filtered by class_id and/or size.
+        """
+
+        data_1, class_id_1, confidence_1 = self._data_1.get(class_id, size_category)
+        data_2, class_id_2, confidence_2 = self._data_2.get(class_id, size_category)
+
+        if self._metric_target == MetricTarget.MASKS:
+            data_1, data_2 = unify_pad_masks_shape(data_1, data_2)
+
+        return (
+            (data_1, class_id_1, confidence_1),
+            (data_2, class_id_2, confidence_2),
+        )
+
+    def get_classes(self) -> Set[int]:
+        """Return all class IDs."""
+        return self._data_1.get_classes() | self._data_2.get_classes()

supervision/metrics/utils/object_size.py

@@ -0,0 +1,155 @@
+from __future__ import annotations
+
+from enum import Enum
+from typing import TYPE_CHECKING
+
+import numpy as np
+import numpy.typing as npt
+
+from supervision.config import ORIENTED_BOX_COORDINATES
+from supervision.metrics.core import MetricTarget
+
+if TYPE_CHECKING:
+    from supervision.detection.core import Detections
+
+SIZE_THRESHOLDS = (32**2, 96**2)
+
+
+class ObjectSizeCategory(Enum):
+    ANY = -1
+    SMALL = 1
+    MEDIUM = 2
+    LARGE = 3
+
+
+def get_object_size_category(
+    data: npt.NDArray, metric_target: MetricTarget
+) -> npt.NDArray[np.int_]:
+    """
+    Get the size category of an object. Distinguish based on the metric target.
+
+    Args:
+        data (np.ndarray): The object data, shaped (N, ...).
+        metric_target (MetricTarget): Determines whether boxes, masks or
+            oriented bounding boxes are used.
+
+    Returns:
+        (np.ndarray) The size category of each object, matching
+        the enum values of ObjectSizeCategory. Shaped (N,).
+    """
+    if metric_target == MetricTarget.BOXES:
+        return get_bbox_size_category(data)
+    if metric_target == MetricTarget.MASKS:
+        return get_mask_size_category(data)
+    if metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+        return get_obb_size_category(data)
+    raise ValueError("Invalid metric type")
+
+
+def get_bbox_size_category(xyxy: npt.NDArray[np.float32]) -> npt.NDArray[np.int_]:
+    """
+    Get the size category of a bounding boxes array.
+
+    Args:
+        xyxy (np.ndarray): The bounding boxes array shaped (N, 4).
+
+    Returns:
+        (np.ndarray) The size category of each bounding box, matching
+        the enum values of ObjectSizeCategory. Shaped (N,).
+    """
+    if len(xyxy.shape) != 2 or xyxy.shape[1] != 4:
+        raise ValueError("Bounding boxes must be shaped (N, 4)")
+
+    width = xyxy[:, 2] - xyxy[:, 0]
+    height = xyxy[:, 3] - xyxy[:, 1]
+    areas = width * height
+
+    result = np.full(areas.shape, ObjectSizeCategory.ANY.value)
+    SM, LG = SIZE_THRESHOLDS
+    result[areas < SM] = ObjectSizeCategory.SMALL.value
+    result[(areas >= SM) & (areas < LG)] = ObjectSizeCategory.MEDIUM.value
+    result[areas >= LG] = ObjectSizeCategory.LARGE.value
+    return result
+
+
+def get_mask_size_category(mask: npt.NDArray[np.bool_]) -> npt.NDArray[np.int_]:
+    """
+    Get the size category of detection masks.
+
+    Args:
+        mask (np.ndarray): The mask array shaped (N, H, W).
+
+    Returns:
+        (np.ndarray) The size category of each mask, matching
+        the enum values of ObjectSizeCategory. Shaped (N,).
+    """
+    if len(mask.shape) != 3:
+        raise ValueError("Masks must be shaped (N, H, W)")
+
+    areas = np.sum(mask, axis=(1, 2))
+
+    result = np.full(areas.shape, ObjectSizeCategory.ANY.value)
+    SM, LG = SIZE_THRESHOLDS
+    result[areas < SM] = ObjectSizeCategory.SMALL.value
+    result[(areas >= SM) & (areas < LG)] = ObjectSizeCategory.MEDIUM.value
+    result[areas >= LG] = ObjectSizeCategory.LARGE.value
+    return result
+
+
+def get_obb_size_category(xyxyxyxy: npt.NDArray[np.float32]) -> npt.NDArray[np.int_]:
+    """
+    Get the size category of a oriented bounding boxes array.
+
+    Args:
+        xyxyxyxy (np.ndarray): The bounding boxes array shaped (N, 8).
+
+    Returns:
+        (np.ndarray) The size category of each bounding box, matching
+        the enum values of ObjectSizeCategory. Shaped (N,).
+    """
+    if len(xyxyxyxy.shape) != 2 or xyxyxyxy.shape[1] != 8:
+        raise ValueError("Oriented bounding boxes must be shaped (N, 8)")
+
+    # Shoelace formula
+    x1, y1, x2, y2, x3, y3, x4, y4 = xyxyxyxy.T
+    areas = 0.5 * np.abs(
+        (x1 * y2 + x2 * y3 + x3 * y4 + x4 * y1)
+        - (x2 * y1 + x3 * y2 + x4 * y3 + x1 * y4)
+    )
+
+    result = np.full(areas.shape, ObjectSizeCategory.ANY.value)
+    SM, LG = SIZE_THRESHOLDS
+    result[areas < SM] = ObjectSizeCategory.SMALL.value
+    result[(areas >= SM) & (areas < LG)] = ObjectSizeCategory.MEDIUM.value
+    result[areas >= LG] = ObjectSizeCategory.LARGE.value
+    return result
+
+
+def get_detection_size_category(
+    detections: Detections, metric_target: MetricTarget = MetricTarget.BOXES
+) -> npt.NDArray[np.int_]:
+    """
+    Get the size category of a detections object.
+
+    Args:
+        xyxyxyxy (np.ndarray): The bounding boxes array shaped (N, 8).
+        metric_target (MetricTarget): Determines whether boxes, masks or
+            oriented bounding boxes are used.
+
+    Returns:
+        (np.ndarray) The size category of each bounding box, matching
+        the enum values of ObjectSizeCategory. Shaped (N,).
+    """
+    if metric_target == MetricTarget.BOXES:
+        return get_bbox_size_category(detections.xyxy)
+    if metric_target == MetricTarget.MASKS:
+        if detections.mask is None:
+            raise ValueError("Detections mask is not available")
+        return get_mask_size_category(detections.mask)
+    if metric_target == MetricTarget.ORIENTED_BOUNDING_BOXES:
+        if detections.data.get(ORIENTED_BOX_COORDINATES) is None:
+            raise ValueError("Detections oriented bounding boxes are not available")
+        return get_obb_size_category(
+            np.array(detections.data[ORIENTED_BOX_COORDINATES])
+        )
+    raise ValueError("Invalid metric type")

supervision/metrics/utils/utils.py

@@ -0,0 +1,70 @@
+from typing import List, Tuple
+
+import numpy as np
+import numpy.typing as npt
+
+
+def ensure_pandas_installed():
+    try:
+        import pandas  # noqa
+    except ImportError:
+        raise ImportError(
+            "`metrics` extra is required to run the function."
+            " Run `pip install 'supervision[metrics]'` or"
+            " `poetry add supervision -E metrics`"
+        )
+
+
+def pad_mask(mask: npt.NDArray, new_shape: Tuple[int, int]) -> npt.NDArray:
+    """Pad a mask to a new shape, inserting zeros on the right and bottom."""
+    if len(mask.shape) != 3:
+        raise ValueError(f"Invalid mask shape: {mask.shape}. Expected: (N, H, W)")
+
+    new_mask = np.pad(
+        mask,
+        (
+            (0, 0),
+            (0, new_shape[0] - mask.shape[1]),
+            (0, new_shape[1] - mask.shape[2]),
+        ),
+        mode="constant",
+        constant_values=0,
+    )
+
+    return new_mask
+
+
+def unify_pad_masks_shape(
+    *masks: npt.NDArray[np.bool_],
+) -> List[npt.NDArray[np.bool_]]:
+    """
+    Given any number of (N, H, W) mask objects, return copies of the
+    same (H, W), padded to the largest dimensions.
+
+    Args:
+        *masks (np.ndarray): The mask arrays to unify. Each shaped (_, H, W),
+
+    Returns:
+        List[np.ndarray]: The masks, padded to the largest dimensions.
+            Each list element shaped (_, H_max, W_max)
+    """
+    new_h = 0
+    new_w = 0
+    for mask in masks:
+        if len(mask.shape) != 3:
+            raise ValueError(f"Invalid mask shape: {mask.shape}. Expected: (N, H, W)")
+
+        _, h, w = mask.shape
+        new_h = max(new_h, h)
+        new_w = max(new_w, w)
+
+    results = []
+    for mask in masks:
+        results.append(pad_mask(mask, (new_h, new_w)))
+
+    return results
+
+
+def len0_like(data: npt.NDArray) -> npt.NDArray:
+    """Create an empty array with the same shape as input, but with 0 rows."""
+    return np.empty((0, *data.shape[1:]), dtype=data.dtype)

supervision/tracker/byte_tracker/core.py

@@ -7,13 +7,13 @@ from supervision.detection.utils import box_iou_batch
 from supervision.tracker.byte_tracker import matching
 from supervision.tracker.byte_tracker.basetrack import BaseTrack, TrackState
 from supervision.tracker.byte_tracker.kalman_filter import KalmanFilter
-from supervision.utils.internal import deprecated_parameter
 
 
 class STrack(BaseTrack):
     shared_kalman = KalmanFilter()
+    _external_count = 0
 
-    def __init__(self, tlwh, score, class_ids):
+    def __init__(self, tlwh, score, class_ids, minimum_consecutive_frames):
         # wait activate
         self._tlwh = np.asarray(tlwh, dtype=np.float32)
         self.kalman_filter = None
@@ -24,6 +24,10 @@ class STrack(BaseTrack):
         self.class_ids = class_ids
         self.tracklet_len = 0
 
+        self.external_track_id = -1
+
+        self.minimum_consecutive_frames = minimum_consecutive_frames
+
     def predict(self):
         mean_state = self.mean.copy()
         if self.state != TrackState.Tracked:
@@ -53,7 +57,7 @@ class STrack(BaseTrack):
     def activate(self, kalman_filter, frame_id):
         """Start a new tracklet"""
         self.kalman_filter = kalman_filter
-        self.track_id = self.next_id()
+        self.internal_track_id = self.next_id()
         self.mean, self.covariance = self.kalman_filter.initiate(
             self.tlwh_to_xyah(self._tlwh)
         )
@@ -62,6 +66,10 @@ class STrack(BaseTrack):
         self.state = TrackState.Tracked
         if frame_id == 1:
             self.is_activated = True
+
+        if self.minimum_consecutive_frames == 1:
+            self.external_track_id = self.next_external_id()
+
         self.frame_id = frame_id
         self.start_frame = frame_id
 
@@ -71,10 +79,10 @@ class STrack(BaseTrack):
         )
         self.tracklet_len = 0
         self.state = TrackState.Tracked
-        self.is_activated = True
+
         self.frame_id = frame_id
         if new_id:
-            self.track_id = self.next_id()
+            self.internal_track_id = self.next_id()
         self.score = new_track.score
 
     def update(self, new_track, frame_id):
@@ -93,7 +101,10 @@ class STrack(BaseTrack):
             self.mean, self.covariance, self.tlwh_to_xyah(new_tlwh)
         )
         self.state = TrackState.Tracked
-        self.is_activated = True
+        if self.tracklet_len == self.minimum_consecutive_frames:
+            self.is_activated = True
+            if self.external_track_id == -1:
+                self.external_track_id = self.next_external_id()
 
         self.score = new_track.score
 
@@ -131,6 +142,15 @@ class STrack(BaseTrack):
     def to_xyah(self):
         return self.tlwh_to_xyah(self.tlwh)
 
+    @staticmethod
+    def next_external_id():
+        STrack._external_count += 1
+        return STrack._external_count
+
+    @staticmethod
+    def reset_external_counter():
+        STrack._external_count = 0
+
     @staticmethod
     def tlbr_to_tlwh(tlbr):
         ret = np.asarray(tlbr).copy()
@@ -144,7 +164,9 @@ class STrack(BaseTrack):
         return ret
 
     def __repr__(self):
-        return "OT_{}_({}-{})".format(self.track_id, self.start_frame, self.end_frame)
+        return "OT_{}_({}-{})".format(
+            self.internal_track_id, self.start_frame, self.end_frame
+        )
 
 
 def detections2boxes(detections: Detections) -> np.ndarray:
@@ -174,50 +196,31 @@ class ByteTrack:
     </video>
 
     Parameters:
-        track_activation_threshold (float, optional): Detection confidence threshold
+        track_activation_threshold (float): Detection confidence threshold
             for track activation. Increasing track_activation_threshold improves accuracy
             and stability but might miss true detections. Decreasing it increases
             completeness but risks introducing noise and instability.
-        lost_track_buffer (int, optional): Number of frames to buffer when a track is lost.
+        lost_track_buffer (int): Number of frames to buffer when a track is lost.
             Increasing lost_track_buffer enhances occlusion handling, significantly
             reducing the likelihood of track fragmentation or disappearance caused
             by brief detection gaps.
-        minimum_matching_threshold (float, optional): Threshold for matching tracks with detections.
+        minimum_matching_threshold (float): Threshold for matching tracks with detections.
             Increasing minimum_matching_threshold improves accuracy but risks fragmentation.
             Decreasing it improves completeness but risks false positives and drift.
-        frame_rate (int, optional): The frame rate of the video.
+        frame_rate (int): The frame rate of the video.
+        minimum_consecutive_frames (int): Number of consecutive frames that an object must
+            be tracked before it is considered a 'valid' track.
+            Increasing minimum_consecutive_frames prevents the creation of accidental tracks from
+            false detection or double detection, but risks missing shorter tracks.
     """  # noqa: E501 // docs
 
-    @deprecated_parameter(
-        old_parameter="track_buffer",
-        new_parameter="lost_track_buffer",
-        map_function=lambda x: x,
-        warning_message="`{old_parameter}` in `{function_name}` is deprecated and will "
-        "be remove in `supervision-0.23.0`. Use '{new_parameter}' "
-        "instead.",
-    )
-    @deprecated_parameter(
-        old_parameter="track_thresh",
-        new_parameter="track_activation_threshold",
-        map_function=lambda x: x,
-        warning_message="`{old_parameter}` in `{function_name}` is deprecated and will "
-        "be remove in `supervision-0.23.0`. Use '{new_parameter}' "
-        "instead.",
-    )
-    @deprecated_parameter(
-        old_parameter="match_thresh",
-        new_parameter="minimum_matching_threshold",
-        map_function=lambda x: x,
-        warning_message="`{old_parameter}` in `{function_name}` is deprecated and will "
-        "be remove in `supervision-0.23.0`. Use '{new_parameter}' "
-        "instead.",
-    )
     def __init__(
         self,
         track_activation_threshold: float = 0.25,
         lost_track_buffer: int = 30,
         minimum_matching_threshold: float = 0.8,
         frame_rate: int = 30,
+        minimum_consecutive_frames: int = 1,
     ):
         self.track_activation_threshold = track_activation_threshold
         self.minimum_matching_threshold = minimum_matching_threshold
@@ -225,6 +228,7 @@ class ByteTrack:
         self.frame_id = 0
         self.det_thresh = self.track_activation_threshold + 0.1
         self.max_time_lost = int(frame_rate / 30.0 * lost_track_buffer)
+        self.minimum_consecutive_frames = minimum_consecutive_frames
         self.kalman_filter = KalmanFilter()
 
         self.tracked_tracks: List[STrack] = []
@@ -247,7 +251,7 @@ class ByteTrack:
             model = YOLO(<MODEL_PATH>)
             tracker = sv.ByteTrack()
 
-            bounding_box_annotator = sv.BoundingBoxAnnotator()
+            box_annotator = sv.BoxAnnotator()
             label_annotator = sv.LabelAnnotator()
 
             def callback(frame: np.ndarray, index: int) -> np.ndarray:
@@ -257,7 +261,7 @@ class ByteTrack:
 
                 labels = [f"#{tracker_id}" for tracker_id in detections.tracker_id]
 
-                annotated_frame = bounding_box_annotator.annotate(
+                annotated_frame = box_annotator.annotate(
                     scene=frame.copy(), detections=detections)
                 annotated_frame = label_annotator.annotate(
                     scene=annotated_frame, detections=detections, labels=labels)
@@ -285,11 +289,14 @@ class ByteTrack:
             matches, _, _ = matching.linear_assignment(iou_costs, 0.5)
             detections.tracker_id = np.full(len(detections), -1, dtype=int)
             for i_detection, i_track in matches:
-                detections.tracker_id[i_detection] = int(tracks[i_track].track_id)
+                detections.tracker_id[i_detection] = int(
+                    tracks[i_track].external_track_id
+                )
 
             return detections[detections.tracker_id != -1]
 
         else:
+            detections = Detections.empty()
             detections.tracker_id = np.array([], dtype=int)
 
             return detections
@@ -308,6 +315,7 @@ class ByteTrack:
         self.lost_tracks: List[STrack] = []
         self.removed_tracks: List[STrack] = []
         BaseTrack.reset_counter()
+        STrack.reset_external_counter()
 
     def update_with_tensors(self, tensors: np.ndarray) -> List[STrack]:
         """
@@ -345,7 +353,7 @@ class ByteTrack:
         if len(dets) > 0:
             """Detections"""
             detections = [
-                STrack(STrack.tlbr_to_tlwh(tlbr), s, c)
+                STrack(STrack.tlbr_to_tlwh(tlbr), s, c, self.minimum_consecutive_frames)
                 for (tlbr, s, c) in zip(dets, scores_keep, class_ids_keep)
             ]
         else:
@@ -387,7 +395,7 @@ class ByteTrack:
         if len(dets_second) > 0:
             """Detections"""
             detections_second = [
-                STrack(STrack.tlbr_to_tlwh(tlbr), s, c)
+                STrack(STrack.tlbr_to_tlwh(tlbr), s, c, self.minimum_consecutive_frames)
                 for (tlbr, s, c) in zip(dets_second, scores_second, class_ids_second)
             ]
         else:
@@ -454,7 +462,7 @@ class ByteTrack:
         self.lost_tracks = sub_tracks(self.lost_tracks, self.tracked_tracks)
         self.lost_tracks.extend(lost_stracks)
         self.lost_tracks = sub_tracks(self.lost_tracks, self.removed_tracks)
-        self.removed_tracks.extend(removed_stracks)
+        self.removed_tracks = removed_stracks
         self.tracked_tracks, self.lost_tracks = remove_duplicate_tracks(
             self.tracked_tracks, self.lost_tracks
         )
@@ -468,22 +476,22 @@ def joint_tracks(
 ) -> List[STrack]:
     """
     Joins two lists of tracks, ensuring that the resulting list does not
-    contain tracks with duplicate track_id values.
+    contain tracks with duplicate internal_track_id values.
 
     Parameters:
-        track_list_a: First list of tracks (with track_id attribute).
-        track_list_b: Second list of tracks (with track_id attribute).
+        track_list_a: First list of tracks (with internal_track_id attribute).
+        track_list_b: Second list of tracks (with internal_track_id attribute).
 
     Returns:
         Combined list of tracks from track_list_a and track_list_b
-            without duplicate track_id values.
+            without duplicate internal_track_id values.
     """
     seen_track_ids = set()
     result = []
 
     for track in track_list_a + track_list_b:
-        if track.track_id not in seen_track_ids:
-            seen_track_ids.add(track.track_id)
+        if track.internal_track_id not in seen_track_ids:
+            seen_track_ids.add(track.internal_track_id)
             result.append(track)
 
     return result
@@ -492,17 +500,17 @@ def joint_tracks(
 def sub_tracks(track_list_a: List, track_list_b: List) -> List[int]:
     """
     Returns a list of tracks from track_list_a after removing any tracks
-    that share the same track_id with tracks in track_list_b.
+    that share the same internal_track_id with tracks in track_list_b.
 
     Parameters:
-        track_list_a: List of tracks (with track_id attribute).
-        track_list_b: List of tracks (with track_id attribute) to
+        track_list_a: List of tracks (with internal_track_id attribute).
+        track_list_b: List of tracks (with internal_track_id attribute) to
             be subtracted from track_list_a.
     Returns:
         List of remaining tracks from track_list_a after subtraction.
     """
-    tracks = {track.track_id: track for track in track_list_a}
-    track_ids_b = {track.track_id for track in track_list_b}
+    tracks = {track.internal_track_id: track for track in track_list_a}
+    track_ids_b = {track.internal_track_id for track in track_list_b}
 
     for track_id in track_ids_b:
         tracks.pop(track_id, None)

supervision/utils/conversion.py

@@ -8,11 +8,13 @@ from PIL import Image
 from supervision.annotators.base import ImageType
 
 
-def convert_for_annotation_method(annotate_func):
+def ensure_cv2_image_for_annotation(annotate_func):
     """
     Decorates `BaseAnnotator.annotate` implementations, converts scene to
     an image type used internally by the annotators, converts back when annotation
     is complete.
+
+    Assumes the annotators modify the scene in-place.
     """
 
     @wraps(annotate_func)
@@ -21,19 +23,22 @@ def convert_for_annotation_method(annotate_func):
             return annotate_func(self, scene, *args, **kwargs)
 
         if isinstance(scene, Image.Image):
-            scene = pillow_to_cv2(scene)
-            annotated = annotate_func(self, scene, *args, **kwargs)
-            return cv2_to_pillow(image=annotated)
+            scene_np = pillow_to_cv2(scene)
+            annotated_np = annotate_func(self, scene_np, *args, **kwargs)
+            scene.paste(cv2_to_pillow(annotated_np))
+            return scene
 
         raise ValueError(f"Unsupported image type: {type(scene)}")
 
     return wrapper
 
 
-def convert_for_image_processing(image_processing_fun):
+def ensure_cv2_image_for_processing(image_processing_fun):
     """
     Decorates image processing functions that accept np.ndarray, converting `image` to
     np.ndarray, converts back when processing is complete.
+
+    Assumes the annotators do NOT modify the scene in-place.
     """
 
     @wraps(image_processing_fun)
@@ -44,13 +49,37 @@ def convert_for_image_processing(image_processing_fun):
         if isinstance(image, Image.Image):
             scene = pillow_to_cv2(image)
             annotated = image_processing_fun(scene, *args, **kwargs)
-            return cv2_to_pillow(image=annotated)
+            return cv2_to_pillow(annotated)
 
         raise ValueError(f"Unsupported image type: {type(image)}")
 
     return wrapper
 
 
+def ensure_pil_image_for_annotation(annotate_func):
+    """
+    Decorates image processing functions that accept np.ndarray, converting `image` to
+    PIL image, converts back when processing is complete.
+
+    Assumes the annotators modify the scene in-place.
+    """
+
+    @wraps(annotate_func)
+    def wrapper(self, scene: ImageType, *args, **kwargs):
+        if isinstance(scene, np.ndarray):
+            scene_pil = cv2_to_pillow(scene)
+            annotated_pil = annotate_func(self, scene_pil, *args, **kwargs)
+            np.copyto(scene, pillow_to_cv2(annotated_pil))
+            return scene
+
+        if isinstance(scene, Image.Image):
+            return annotate_func(self, scene, *args, **kwargs)
+
+        raise ValueError(f"Unsupported image type: {type(scene)}")
+
+    return wrapper
+
+
 def images_to_cv2(images: List[ImageType]) -> List[np.ndarray]:
     """
     Converts images provided either as Pillow images or OpenCV
@@ -67,7 +96,7 @@ def images_to_cv2(images: List[ImageType]) -> List[np.ndarray]:
     result = []
     for image in images:
         if issubclass(type(image), Image.Image):
-            image = pillow_to_cv2(image=image)
+            image = pillow_to_cv2(image)
         result.append(image)
     return result