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
@@ -19,14 +18,12 @@ 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: '1.7.9'
       hooks:
@@ -34,18 +31,8 @@ repos:
           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: v0.5.1
+      rev: v0.6.3
       hooks:
       -   id: ruff
           args: [--fix, --exit-non-zero-on-fix]

CONTRIBUTING.md

@@ -173,7 +173,7 @@ Furthermore, we have integrated a pre-commit GitHub Action into our workflow. Th
 
 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
+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.
 

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)
@@ -110,18 +111,21 @@ Supervision provides a set of [utils](https://supervision.roboflow.com/latest/da
 
 ```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>
@@ -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%"

demo.ipynb

@@ -972,7 +972,7 @@
     }
    ],
    "source": [
-    "from supervision.assets import download_assets, VideoAssets\n",
+    "from supervision.assets import VideoAssets, download_assets\n",
     "\n",
     "download_assets(VideoAssets.VEHICLES)\n",
     "VIDEO_PATH = VideoAssets.VEHICLES.value"

docs/changelog.md

@@ -1,3 +1,136 @@
+### 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).
@@ -360,7 +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.
 

docs/datasets/core.md

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

docs/deprecated.md

@@ -7,15 +7,24 @@ status: deprecated
 
 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 `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) 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) will be removed in `supervision-0.24.0`.
-- Constructing `DetectionDataset` and `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.
+
+- 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` has been removed. `BoundingBoxAnnotator` will be removed in `supervision-0.26.0`.
+- `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.

docs/detection/annotators.md

@@ -303,7 +303,7 @@ status: new
         font_path="<TTF_FONT_PATH>",
         text_position=sv.Position.CENTER
     )
-    annotated_frame = label_annotator.annotate(
+    annotated_frame = rich_label_annotator.annotate(
         scene=image.copy(),
         detections=detections,
         labels=labels
@@ -316,6 +316,33 @@ status: new
 
     </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
@@ -384,7 +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:
@@ -415,7 +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:
@@ -433,6 +460,27 @@ 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.BoxAnnotator">BoxAnnotator</a></h2>
 </div>
@@ -529,6 +577,12 @@ status: new
 
 :::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>
@@ -553,6 +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/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/utils.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Detection Utils
@@ -78,19 +77,19 @@ status: new
 :::supervision.detection.utils.pad_boxes
 
 <div class="md-typeset">
-  <h2><a href="#supervision.detection.utils.contains_holes">contains_holes</a></h2>
+  <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.xywh_to_xyxy">xywh_to_xyxy</a></h2>
+  <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.xcycwh_to_xyxy">xcycwh_to_xyxy</a></h2>
+  <h2><a href="#supervision.detection.utils.contains_holes">contains_holes</a></h2>
 </div>
 
 :::supervision.detection.utils.contains_holes

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/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/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

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

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/notebooks/serialise-detections-to-csv.ipynb

@@ -10,7 +10,7 @@
                 "\n",
                 "---\n",
                 "\n",
-                "[![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow/supervision/blob/develop/docs/notebooks/detections-to-jsonsink.ipynb)\n",
+                "[![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow/supervision/blob/develop/docs/notebooks/serialise-detections-to-csv.ipynb)\n",
                 "\n",
                 "This cookbook introduce [sv.CSVSink](https://supervision.roboflow.com/latest/detection/tools/save_detections/#supervision.detection.tools.csv_sink.CSVSink) tool designed to write captured object detection data to file from video streams/file"
             ]

docs/notebooks/serialise-detections-to-json.ipynb

@@ -10,7 +10,7 @@
                 "\n",
                 "---\n",
                 "\n",
-                "[![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow/supervision/blob/develop/docs/notebooks/detections-to-jsonsink.ipynb)\n",
+                "[![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow/supervision/blob/develop/docs/notebooks/serialise-detections-to-json.ipynb)\n",
                 "\n",
                 "This cookbook introduce [sv.JSONSink](https://supervision.roboflow.com/develop/detection/tools/save_detections/#supervision.detection.tools.json_sink.JSONSink) tool designed to write captured object detection data to file from video streams/file"
             ]

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

@@ -11,17 +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>
-      <p class="card repo-card" data-url="/develop/notebooks/serialise-detections-to-csv" data-name="Serialise Detections to a CSV File"  data-labels="DETECTIONS,CSV SINK,INFERENCE" data-version="v0.21.0" data-author="onuralpszr"></p>
-      <p class="card repo-card" data-url="/develop/notebooks/serialise-detections-to-json" data-name="Serialise Detections to a JSON File"  data-labels="DETECTIONS,JSON SINK,INFERENCE" data-version="v0.21.0" data-author="onuralpszr"></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/utils/file.md

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

docs/utils/image.md

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

docs/utils/video.md

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

examples/count_people_in_zone/inference_example.py

@@ -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/ultralytics_example.py

@@ -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

@@ -95,7 +95,7 @@ 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_optimal_line_thickness(

examples/speed_estimation/ultralytics_example.py

@@ -73,7 +73,7 @@ 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_optimal_line_thickness(

examples/speed_estimation/yolo_nas_example.py

@@ -74,7 +74,7 @@ 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_optimal_line_thickness(

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

mkdocs.yml

@@ -33,7 +33,6 @@ extra_css:
   - stylesheets/extra.css
   - stylesheets/cookbooks-card.css
 
-
 nav:
   - Supervision: index.md
   - Learn:
@@ -42,52 +41,53 @@ nav:
       - Filter Detections: how_to/filter_detections.md
       - Detect Small Objects: how_to/detect_small_objects.md
       - Track Objects on Video: how_to/track_objects.md
+      - Process Datasets: how_to/process_datasets.md
 
   - Reference - Code API:
-    - Detection and Segmentation:
-        - Core: detection/core.md
-        - Annotators: detection/annotators.md
-        - Metrics: detection/metrics.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
-    - 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
+      - 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
+      - Contributing: contributing.md
+      - Code of Conduct: code_of_conduct.md
+      - License: license.md
   - Release Notes:
-        - Changelog: changelog.md
-        - Deprecated: deprecated.md
-
-
+      - Changelog: changelog.md
+      - Deprecated: deprecated.md
 
 theme:
-  name: 'material'
+  name: "material"
   icon:
     edit: material/pencil
   logo: assets/supervision-lenny.png
@@ -104,19 +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
@@ -127,10 +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:
@@ -154,8 +153,6 @@ plugins:
   - git-revision-date-localized:
       enable_creation_date: true
 
-
-
 markdown_extensions:
   - admonition
   - pymdownx.details

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "supervision"
-version = "0.22.0"
+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>"]
@@ -40,7 +40,10 @@ classifiers = [
 
 [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" },
@@ -48,20 +51,24 @@ scipy = [
 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,<=2.32.3", optional = true }
-tqdm = { version = ">=4.62.3,<=4.66.4", optional = true }
-pillow = ">=9.4"
+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"]
+metrics = ["pandas", "pandas-stubs"]
 
 [tool.poetry.group.dev.dependencies]
 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"
@@ -76,10 +83,9 @@ docutils = [
     { 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,<0.26" }
+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
@@ -87,9 +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.isort]
-line_length = 88
-profile = "black"
+[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.bandit]
 target = ["test", "supervision"]
@@ -154,7 +166,7 @@ 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 = [
@@ -223,6 +235,10 @@ convention = "google"
 # 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"

release_process.md

@@ -9,10 +9,15 @@ It assumes you already have the code changes, as well as a draft of the release
    - 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.
+   - 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.

supervision/__init__.py

@@ -7,6 +7,7 @@ except importlib_metadata.PackageNotFoundError:
     __version__ = "development"
 
 from supervision.annotators.core import (
+    BackgroundOverlayAnnotator,
     BlurAnnotator,
     BoundingBoxAnnotator,
     BoxAnnotator,
@@ -18,6 +19,7 @@ from supervision.annotators.core import (
     EllipseAnnotator,
     HaloAnnotator,
     HeatMapAnnotator,
+    IconAnnotator,
     LabelAnnotator,
     MaskAnnotator,
     OrientedBoxAnnotator,

supervision/annotators/core.py

@@ -1,9 +1,11 @@
+from functools import lru_cache
 from math import sqrt
 from typing import List, Optional, Tuple, Union
 
 import cv2
 import numpy as np
-from PIL import ImageDraw, ImageFont
+import numpy.typing as npt
+from PIL import Image, ImageDraw, ImageFont
 
 from supervision.annotators.base import BaseAnnotator, ImageType
 from supervision.annotators.utils import (
@@ -22,7 +24,12 @@ from supervision.utils.conversion import (
     ensure_cv2_image_for_annotation,
     ensure_pil_image_for_annotation,
 )
-from supervision.utils.image import crop_image, overlay_image, scale_image
+from supervision.utils.image import (
+    crop_image,
+    letterbox_image,
+    overlay_image,
+    scale_image,
+)
 from supervision.utils.internal import deprecated
 
 
@@ -42,7 +49,7 @@ class BoxAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             thickness (int): Thickness of the bounding box lines.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -87,6 +94,7 @@ class BoxAnnotator(BaseAnnotator):
         ![bounding-box-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/bounding-box-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
             color = resolve_color(
@@ -127,7 +135,7 @@ class BoundingBoxAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             thickness (int): Thickness of the bounding box lines.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -172,6 +180,7 @@ class BoundingBoxAnnotator(BaseAnnotator):
         ![bounding-box-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/bounding-box-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
             color = resolve_color(
@@ -208,7 +217,7 @@ class OrientedBoxAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             thickness (int): Thickness of the bounding box lines.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -256,12 +265,13 @@ class OrientedBoxAnnotator(BaseAnnotator):
             )
             ```
         """  # noqa E501 // docs
-
+        assert isinstance(scene, np.ndarray)
         if detections.data is None or ORIENTED_BOX_COORDINATES not in detections.data:
             return scene
+        obb_boxes = np.array(detections.data[ORIENTED_BOX_COORDINATES]).astype(int)
 
         for detection_idx in range(len(detections)):
-            bbox = np.intp(detections.data.get(ORIENTED_BOX_COORDINATES)[detection_idx])
+            obb = obb_boxes[detection_idx]
             color = resolve_color(
                 color=self.color,
                 detections=detections,
@@ -271,7 +281,7 @@ class OrientedBoxAnnotator(BaseAnnotator):
                 else custom_color_lookup,
             )
 
-            cv2.drawContours(scene, [bbox], 0, color.as_bgr(), self.thickness)
+            cv2.drawContours(scene, [obb], 0, color.as_bgr(), self.thickness)
 
         return scene
 
@@ -296,7 +306,7 @@ class MaskAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             opacity (float): Opacity of the overlay mask. Must be between `0` and `1`.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -342,6 +352,7 @@ class MaskAnnotator(BaseAnnotator):
         ![mask-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/mask-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         if detections.mask is None:
             return scene
 
@@ -385,7 +396,7 @@ class PolygonAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             thickness (int): Thickness of the polygon lines.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -431,6 +442,7 @@ class PolygonAnnotator(BaseAnnotator):
         ![polygon-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/polygon-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         if detections.mask is None:
             return scene
 
@@ -471,7 +483,7 @@ class ColorAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             opacity (float): Opacity of the overlay mask. Must be between `0` and `1`.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -517,6 +529,7 @@ class ColorAnnotator(BaseAnnotator):
         ![box-mask-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/box-mask-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         scene_with_boxes = scene.copy()
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
@@ -565,7 +578,7 @@ class HaloAnnotator(BaseAnnotator):
             opacity (float): Opacity of the overlay mask. Must be between `0` and `1`.
             kernel_size (int): The size of the average pooling kernel used for creating
                 the halo.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -612,6 +625,7 @@ class HaloAnnotator(BaseAnnotator):
         ![halo-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/halo-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         if detections.mask is None:
             return scene
         colored_mask = np.zeros_like(scene, dtype=np.uint8)
@@ -663,7 +677,7 @@ class EllipseAnnotator(BaseAnnotator):
             thickness (int): Thickness of the ellipse lines.
             start_angle (int): Starting angle of the ellipse.
             end_angle (int): Ending angle of the ellipse.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -711,6 +725,7 @@ class EllipseAnnotator(BaseAnnotator):
         ![ellipse-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/ellipse-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
             color = resolve_color(
@@ -755,7 +770,7 @@ class BoxCornerAnnotator(BaseAnnotator):
                 annotating detections.
             thickness (int): Thickness of the corner lines.
             corner_length (int): Length of each corner line.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -802,6 +817,7 @@ class BoxCornerAnnotator(BaseAnnotator):
         ![box-corner-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/box-corner-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
             color = resolve_color(
@@ -843,7 +859,7 @@ class CircleAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             thickness (int): Thickness of the circle line.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
 
@@ -891,6 +907,7 @@ class CircleAnnotator(BaseAnnotator):
         ![circle-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/circle-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
             center = ((x1 + x2) // 2, (y1 + y2) // 2)
@@ -927,6 +944,7 @@ class DotAnnotator(BaseAnnotator):
         position: Position = Position.CENTER,
         color_lookup: ColorLookup = ColorLookup.CLASS,
         outline_thickness: int = 0,
+        outline_color: Union[Color, ColorPalette] = Color.BLACK,
     ):
         """
         Args:
@@ -937,12 +955,16 @@ class DotAnnotator(BaseAnnotator):
             color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
             outline_thickness (int): Thickness of the outline of the dot.
+            outline_color (Union[Color, ColorPalette]): The color or color palette to
+                use for outline. It is activated by setting outline_thickness to a value
+                greater than 0.
         """
         self.color: Union[Color, ColorPalette] = color
         self.radius: int = radius
         self.position: Position = position
         self.color_lookup: ColorLookup = color_lookup
         self.outline_thickness = outline_thickness
+        self.outline_color: Union[Color, ColorPalette] = outline_color
 
     @ensure_cv2_image_for_annotation
     def annotate(
@@ -983,6 +1005,7 @@ class DotAnnotator(BaseAnnotator):
         ![dot-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/dot-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         xy = detections.get_anchors_coordinates(anchor=self.position)
         for detection_idx in range(len(detections)):
             color = resolve_color(
@@ -997,8 +1020,20 @@ class DotAnnotator(BaseAnnotator):
 
             cv2.circle(scene, center, self.radius, color.as_bgr(), -1)
             if self.outline_thickness:
+                outline_color = resolve_color(
+                    color=self.outline_color,
+                    detections=detections,
+                    detection_idx=detection_idx,
+                    color_lookup=self.color_lookup
+                    if custom_color_lookup is None
+                    else custom_color_lookup,
+                )
                 cv2.circle(
-                    scene, center, self.radius, (0, 0, 0), self.outline_thickness
+                    scene,
+                    center,
+                    self.radius,
+                    outline_color.as_bgr(),
+                    self.outline_thickness,
                 )
         return scene
 
@@ -1011,7 +1046,7 @@ class LabelAnnotator(BaseAnnotator):
     def __init__(
         self,
         color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
-        text_color: Color = Color.WHITE,
+        text_color: Union[Color, ColorPalette] = Color.WHITE,
         text_scale: float = 0.5,
         text_thickness: int = 1,
         text_padding: int = 10,
@@ -1023,20 +1058,21 @@ class LabelAnnotator(BaseAnnotator):
         Args:
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating the text background.
-            text_color (Color): The color to use for the text.
+            text_color (Union[Color, ColorPalette]): The color or color palette to use
+                for the text.
             text_scale (float): Font scale for the text.
             text_thickness (int): Thickness of the text characters.
             text_padding (int): Padding around the text within its background box.
             text_position (Position): Position of the text relative to the detection.
                 Possible values are defined in the `Position` enum.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
             border_radius (int): The radius to apply round edges. If the selected
                 value is higher than the lower dimension, width or height, is clipped.
         """
         self.border_radius: int = border_radius
         self.color: Union[Color, ColorPalette] = color
-        self.text_color: Color = text_color
+        self.text_color: Union[Color, ColorPalette] = text_color
         self.text_scale: float = text_scale
         self.text_thickness: int = text_thickness
         self.text_padding: int = text_padding
@@ -1091,6 +1127,7 @@ class LabelAnnotator(BaseAnnotator):
         ![label-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/label-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         font = cv2.FONT_HERSHEY_SIMPLEX
         anchors_coordinates = detections.get_anchors_coordinates(
             anchor=self.text_anchor
@@ -1114,10 +1151,21 @@ class LabelAnnotator(BaseAnnotator):
                 ),
             )
 
+            text_color = resolve_color(
+                color=self.text_color,
+                detections=detections,
+                detection_idx=detection_idx,
+                color_lookup=(
+                    self.color_lookup
+                    if custom_color_lookup is None
+                    else custom_color_lookup
+                ),
+            )
+
             if labels is not None:
                 text = labels[detection_idx]
-            elif detections[CLASS_NAME_DATA_FIELD] is not None:
-                text = detections[CLASS_NAME_DATA_FIELD][detection_idx]
+            elif CLASS_NAME_DATA_FIELD in detections.data:
+                text = detections.data[CLASS_NAME_DATA_FIELD][detection_idx]
             elif detections.class_id is not None:
                 text = str(detections.class_id[detection_idx])
             else:
@@ -1152,7 +1200,7 @@ class LabelAnnotator(BaseAnnotator):
                 org=(text_x, text_y),
                 fontFace=font,
                 fontScale=self.text_scale,
-                color=self.text_color.as_rgb(),
+                color=text_color.as_bgr(),
                 thickness=self.text_thickness,
                 lineType=cv2.LINE_AA,
             )
@@ -1210,7 +1258,7 @@ class RichLabelAnnotator(BaseAnnotator):
     def __init__(
         self,
         color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
-        text_color: Color = Color.WHITE,
+        text_color: Union[Color, ColorPalette] = Color.WHITE,
         font_path: Optional[str] = None,
         font_size: int = 10,
         text_padding: int = 10,
@@ -1222,7 +1270,7 @@ class RichLabelAnnotator(BaseAnnotator):
         Args:
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating the text background.
-            text_color (Color): The color to use for the text.
+            text_color (Union[Color, ColorPalette]): The color to use for the text.
             font_path (Optional[str]): Path to the font file (e.g., ".ttf" or ".otf")
                 to use for rendering text. If `None`, the default PIL font will be used.
             font_size (int): Font size for the text.
@@ -1296,6 +1344,7 @@ class RichLabelAnnotator(BaseAnnotator):
             ```
 
         """
+        assert isinstance(scene, Image.Image)
         draw = ImageDraw.Draw(scene)
         anchors_coordinates = detections.get_anchors_coordinates(
             anchor=self.text_anchor
@@ -1317,10 +1366,22 @@ class RichLabelAnnotator(BaseAnnotator):
                     else custom_color_lookup
                 ),
             )
+
+            text_color = resolve_color(
+                color=self.text_color,
+                detections=detections,
+                detection_idx=detection_idx,
+                color_lookup=(
+                    self.color_lookup
+                    if custom_color_lookup is None
+                    else custom_color_lookup
+                ),
+            )
+
             if labels is not None:
                 text = labels[detection_idx]
-            elif detections[CLASS_NAME_DATA_FIELD] is not None:
-                text = detections[CLASS_NAME_DATA_FIELD][detection_idx]
+            elif CLASS_NAME_DATA_FIELD in detections.data:
+                text = detections.data[CLASS_NAME_DATA_FIELD][detection_idx]
             elif detections.class_id is not None:
                 text = str(detections.class_id[detection_idx])
             else:
@@ -1350,7 +1411,7 @@ class RichLabelAnnotator(BaseAnnotator):
                 xy=(text_x, text_y),
                 text=text,
                 font=self.font,
-                fill=self.text_color.as_rgb(),
+                fill=text_color.as_rgb(),
             )
         return scene
 
@@ -1368,6 +1429,106 @@ class RichLabelAnnotator(BaseAnnotator):
         return font
 
 
+class IconAnnotator(BaseAnnotator):
+    """
+    A class for drawing an icon on an image, using provided detections.
+    """
+
+    def __init__(
+        self,
+        icon_resolution_wh: Tuple[int, int] = (64, 64),
+        icon_position: Position = Position.TOP_CENTER,
+        offset_xy: Tuple[int, int] = (0, 0),
+    ):
+        """
+        Args:
+            icon_resolution_wh (Tuple[int, int]): The size of drawn icons.
+                All icons will be resized to this resolution, keeping the aspect ratio.
+            icon_position (Position): The position of the icon.
+            offset_xy (Tuple[int, int]): The offset to apply to the icon position,
+                in pixels. Can be both positive and negative.
+        """
+        self.icon_resolution_wh = icon_resolution_wh
+        self.position = icon_position
+        self.offset_xy = offset_xy
+
+    @ensure_cv2_image_for_annotation
+    def annotate(
+        self, scene: ImageType, detections: Detections, icon_path: Union[str, List[str]]
+    ) -> ImageType:
+        """
+        Annotates the given scene with given icons.
+
+        Args:
+            scene (ImageType): The image where labels will be drawn.
+                `ImageType` is a flexible type, accepting either `numpy.ndarray`
+                or `PIL.Image.Image`.
+            detections (Detections): Object detections to annotate.
+            icon_path (Union[str, List[str]]): The path to the PNG image to use as an
+                icon. Must be a single path or a list of paths, one for each detection.
+                Pass an empty string `""` to draw nothing.
+
+        Returns:
+            The annotated image, matching the type of `scene` (`numpy.ndarray`
+                or `PIL.Image.Image`)
+
+        Example:
+            ```python
+            import supervision as sv
+
+            image = ...
+            detections = sv.Detections(...)
+
+            available_icons = ["roboflow.png", "lenny.png"]
+            icon_paths = [np.random.choice(available_icons) for _ in detections]
+
+            icon_annotator = sv.IconAnnotator()
+            annotated_frame = icon_annotator.annotate(
+                scene=image.copy(),
+                detections=detections,
+                icon_path=icon_paths
+            )
+            ```
+
+        ![icon-annotator-example](https://media.roboflow.com/
+        supervision-annotator-examples/icon-annotator-example.png)
+        """
+        assert isinstance(scene, np.ndarray)
+        if isinstance(icon_path, list) and len(icon_path) != len(detections):
+            raise ValueError(
+                f"The number of icon paths provided ({len(icon_path)}) does not match "
+                f"the number of detections ({len(detections)}). Either provide a single"
+                f" icon path or one for each detection."
+            )
+
+        xy = detections.get_anchors_coordinates(anchor=self.position).astype(int)
+
+        for detection_idx in range(len(detections)):
+            current_path = (
+                icon_path if isinstance(icon_path, str) else icon_path[detection_idx]
+            )
+            if current_path == "":
+                continue
+            icon = self._load_icon(current_path)
+            icon_h, icon_w = icon.shape[:2]
+
+            x = int(xy[detection_idx, 0] - icon_w / 2 + self.offset_xy[0])
+            y = int(xy[detection_idx, 1] - icon_h / 2 + self.offset_xy[1])
+
+            scene[:] = overlay_image(scene, icon, (x, y))
+        return scene
+
+    @lru_cache
+    def _load_icon(self, icon_path: str) -> np.ndarray:
+        icon = cv2.imread(icon_path, cv2.IMREAD_UNCHANGED)
+        if icon is None:
+            raise FileNotFoundError(
+                f"Error: Couldn't load the icon image from {icon_path}"
+            )
+        icon = letterbox_image(image=icon, resolution_wh=self.icon_resolution_wh)
+        return icon
+
+
 class BlurAnnotator(BaseAnnotator):
     """
     A class for blurring regions in an image using provided detections.
@@ -1416,6 +1577,7 @@ class BlurAnnotator(BaseAnnotator):
         ![blur-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/blur-annotator-example-purple.png)
         """
+        assert isinstance(scene, np.ndarray)
         image_height, image_width = scene.shape[:2]
         clipped_xyxy = clip_boxes(
             xyxy=detections.xyxy, resolution_wh=(image_width, image_height)
@@ -1457,7 +1619,7 @@ class TraceAnnotator(BaseAnnotator):
             trace_length (int): The maximum length of the trace in terms of historical
                 points. Defaults to `30`.
             thickness (int): The thickness of the trace lines. Defaults to `2`.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.color: Union[Color, ColorPalette] = color
@@ -1514,8 +1676,14 @@ class TraceAnnotator(BaseAnnotator):
         ![trace-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/trace-annotator-example-purple.png)
         """
-        self.trace.put(detections)
+        assert isinstance(scene, np.ndarray)
+        if detections.tracker_id is None:
+            raise ValueError(
+                "The `tracker_id` field is missing in the provided detections."
+                " See more: https://supervision.roboflow.com/latest/how_to/track_objects"
+            )
 
+        self.trace.put(detections)
         for detection_idx in range(len(detections)):
             tracker_id = int(detections.tracker_id[detection_idx])
             color = resolve_color(
@@ -1568,9 +1736,9 @@ class HeatMapAnnotator(BaseAnnotator):
         self.opacity = opacity
         self.radius = radius
         self.kernel_size = kernel_size
-        self.heat_mask = None
         self.top_hue = top_hue
         self.low_hue = low_hue
+        self.heat_mask: Optional[npt.NDArray[np.float32]] = None
 
     @ensure_cv2_image_for_annotation
     def annotate(self, scene: ImageType, detections: Detections) -> ImageType:
@@ -1597,7 +1765,7 @@ class HeatMapAnnotator(BaseAnnotator):
             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:
@@ -1612,12 +1780,20 @@ class HeatMapAnnotator(BaseAnnotator):
         ![heatmap-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/heat-map-annotator-example-purple.png)
         """
-
+        assert isinstance(scene, np.ndarray)
         if self.heat_mask is None:
-            self.heat_mask = np.zeros(scene.shape[:2])
+            self.heat_mask = np.zeros(scene.shape[:2], dtype=np.float32)
+
         mask = np.zeros(scene.shape[:2])
         for xy in detections.get_anchors_coordinates(self.position):
-            cv2.circle(mask, (int(xy[0]), int(xy[1])), self.radius, 1, -1)
+            x, y = int(xy[0]), int(xy[1])
+            cv2.circle(
+                img=mask,
+                center=(x, y),
+                radius=self.radius,
+                color=(1,),
+                thickness=-1,  # fill
+            )
         self.heat_mask = mask + self.heat_mask
         temp = self.heat_mask.copy()
         temp = self.low_hue - temp / temp.max() * (self.low_hue - self.top_hue)
@@ -1685,6 +1861,7 @@ class PixelateAnnotator(BaseAnnotator):
         ![pixelate-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/pixelate-annotator-example-10.png)
         """
+        assert isinstance(scene, np.ndarray)
         image_height, image_width = scene.shape[:2]
         clipped_xyxy = clip_boxes(
             xyxy=detections.xyxy, resolution_wh=(image_width, image_height)
@@ -1720,6 +1897,7 @@ class TriangleAnnotator(BaseAnnotator):
         position: Position = Position.TOP_CENTER,
         color_lookup: ColorLookup = ColorLookup.CLASS,
         outline_thickness: int = 0,
+        outline_color: Union[Color, ColorPalette] = Color.BLACK,
     ):
         """
         Args:
@@ -1731,6 +1909,9 @@ class TriangleAnnotator(BaseAnnotator):
             color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
             outline_thickness (int): Thickness of the outline of the triangle.
+            outline_color (Union[Color, ColorPalette]): The color or color palette to
+                use for outline. It is activated by setting outline_thickness to a value
+                greater than 0.
         """
         self.color: Union[Color, ColorPalette] = color
         self.base: int = base
@@ -1738,6 +1919,7 @@ class TriangleAnnotator(BaseAnnotator):
         self.position: Position = position
         self.color_lookup: ColorLookup = color_lookup
         self.outline_thickness: int = outline_thickness
+        self.outline_color: Union[Color, ColorPalette] = outline_color
 
     @ensure_cv2_image_for_annotation
     def annotate(
@@ -1778,6 +1960,7 @@ class TriangleAnnotator(BaseAnnotator):
         ![triangle-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/triangle-annotator-example.png)
         """
+        assert isinstance(scene, np.ndarray)
         xy = detections.get_anchors_coordinates(anchor=self.position)
         for detection_idx in range(len(detections)):
             color = resolve_color(
@@ -1800,8 +1983,20 @@ class TriangleAnnotator(BaseAnnotator):
 
             cv2.fillPoly(scene, [vertices], color.as_bgr())
             if self.outline_thickness:
+                outline_color = resolve_color(
+                    color=self.outline_color,
+                    detections=detections,
+                    detection_idx=detection_idx,
+                    color_lookup=self.color_lookup
+                    if custom_color_lookup is None
+                    else custom_color_lookup,
+                )
                 cv2.polylines(
-                    scene, [vertices], True, (0, 0, 0), thickness=self.outline_thickness
+                    scene,
+                    [vertices],
+                    True,
+                    outline_color.as_bgr(),
+                    thickness=self.outline_thickness,
                 )
         return scene
 
@@ -1824,7 +2019,7 @@ class RoundBoxAnnotator(BaseAnnotator):
             color (Union[Color, ColorPalette]): The color or color palette to use for
                 annotating detections.
             thickness (int): Thickness of the bounding box lines.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
             roundness (float): Percent of roundness for edges of bounding box.
                 Value must be float 0 < roundness <= 1.0
@@ -1878,7 +2073,7 @@ class RoundBoxAnnotator(BaseAnnotator):
         ![round-box-annotator-example](https://media.roboflow.com/
         supervision-annotator-examples/round-box-annotator-example-purple.png)
         """
-
+        assert isinstance(scene, np.ndarray)
         for detection_idx in range(len(detections)):
             x1, y1, x2, y2 = detections.xyxy[detection_idx].astype(int)
             color = resolve_color(
@@ -1951,7 +2146,7 @@ class PercentageBarAnnotator(BaseAnnotator):
         border_color: Color = Color.BLACK,
         position: Position = Position.TOP_CENTER,
         color_lookup: ColorLookup = ColorLookup.CLASS,
-        border_thickness: int = None,
+        border_thickness: Optional[int] = None,
     ):
         """
         Args:
@@ -1961,9 +2156,9 @@ class PercentageBarAnnotator(BaseAnnotator):
                 annotating detections.
             border_color (Color): The color of the border lines.
             position (Position): The anchor position of drawing the percentage bar.
-            color_lookup (str): Strategy for mapping colors to annotations.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
                 Options are `INDEX`, `CLASS`, `TRACK`.
-            border_thickness (int): The thickness of the border lines.
+            border_thickness (Optional[int]): The thickness of the border lines.
         """
         self.height: int = height
         self.width: int = width
@@ -2022,9 +2217,9 @@ class PercentageBarAnnotator(BaseAnnotator):
         ![percentage-bar-example](https://media.roboflow.com/
         supervision-annotator-examples/percentage-bar-annotator-example-purple.png)
         """
-        self.validate_custom_values(
-            custom_values=custom_values, detections_count=len(detections)
-        )
+        assert isinstance(scene, np.ndarray)
+        self.validate_custom_values(custom_values=custom_values, detections=detections)
+
         anchors = detections.get_anchors_coordinates(anchor=self.position)
         for detection_idx in range(len(detections)):
             anchor = anchors[detection_idx]
@@ -2035,11 +2230,11 @@ class PercentageBarAnnotator(BaseAnnotator):
             )
             border_width = border_coordinates[1][0] - border_coordinates[0][0]
 
-            value = (
-                custom_values[detection_idx]
-                if custom_values is not None
-                else detections.confidence[detection_idx]
-            )
+            if custom_values is not None:
+                value = custom_values[detection_idx]
+            else:
+                assert detections.confidence is not None  # MyPy type hint
+                value = detections.confidence[detection_idx]
 
             color = resolve_color(
                 color=self.color,
@@ -2099,15 +2294,23 @@ class PercentageBarAnnotator(BaseAnnotator):
 
     @staticmethod
     def validate_custom_values(
-        custom_values: Optional[Union[np.ndarray, List[float]]], detections_count: int
+        custom_values: Optional[Union[np.ndarray, List[float]]], detections: Detections
     ) -> None:
-        if custom_values is not None:
+        if custom_values is None:
+            if detections.confidence is None:
+                raise ValueError(
+                    "The provided detections do not contain confidence values. "
+                    "Please provide `custom_values` or ensure that the detections "
+                    "contain confidence values (e.g. by using a different model)."
+                )
+
+        else:
             if not isinstance(custom_values, (np.ndarray, list)):
                 raise TypeError(
                     "custom_values must be either a numpy array or a list of floats."
                 )
 
-            if len(custom_values) != detections_count:
+            if len(custom_values) != len(detections):
                 raise ValueError(
                     "The length of custom_values must match the number of detections."
                 )
@@ -2124,7 +2327,7 @@ class CropAnnotator(BaseAnnotator):
     def __init__(
         self,
         position: Position = Position.TOP_CENTER,
-        scale_factor: int = 2,
+        scale_factor: float = 2.0,
         border_color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
         border_thickness: int = 2,
         border_color_lookup: ColorLookup = ColorLookup.CLASS,
@@ -2133,7 +2336,7 @@ class CropAnnotator(BaseAnnotator):
         Args:
             position (Position): The anchor position for placing the cropped and scaled
                 part of the detection in the scene.
-            scale_factor (int): The factor by which to scale the cropped image part. A
+            scale_factor (float): The factor by which to scale the cropped image part. A
                 factor of 2, for example, would double the size of the cropped area,
                 allowing for a closer view of the detection.
             border_color (Union[Color, ColorPalette]): The color or color palette to
@@ -2143,7 +2346,7 @@ class CropAnnotator(BaseAnnotator):
                 annotations. Options are `INDEX`, `CLASS`, `TRACK`.
         """
         self.position: Position = position
-        self.scale_factor: int = scale_factor
+        self.scale_factor: float = scale_factor
         self.border_color: Union[Color, ColorPalette] = border_color
         self.border_thickness: int = border_thickness
         self.border_color_lookup: ColorLookup = border_color_lookup
@@ -2187,6 +2390,7 @@ class CropAnnotator(BaseAnnotator):
             )
             ```
         """
+        assert isinstance(scene, np.ndarray)
         crops = [
             crop_image(image=scene, xyxy=xyxy) for xyxy in detections.xyxy.astype(int)
         ]
@@ -2259,3 +2463,85 @@ class CropAnnotator(BaseAnnotator):
             )
         elif position == Position.BOTTOM_RIGHT:
             return (anchor_x, anchor_y), (anchor_x + width, anchor_y + height)
+
+
+class BackgroundOverlayAnnotator(BaseAnnotator):
+    """
+    A class for drawing a colored overlay on the background of an image outside
+    the region of detections.
+
+    If masks are provided, the background is colored outside the masks.
+    If masks are not provided, the background is colored outside the bounding boxes.
+
+    You can use the `force_box` parameter to force the annotator to use bounding boxes.
+
+    !!! warning
+
+        This annotator uses `sv.Detections.mask`.
+    """
+
+    def __init__(
+        self,
+        color: Color = Color.BLACK,
+        opacity: float = 0.5,
+        force_box: bool = False,
+    ):
+        """
+        Args:
+            color (Color): The color to use for annotating detections.
+            opacity (float): Opacity of the overlay mask. Must be between `0` and `1`.
+            force_box (bool): If `True`, forces the annotator to use bounding boxes when
+                masks are provided in the supplied sv.Detections.
+        """
+        self.color: Color = color
+        self.opacity = opacity
+        self.force_box = force_box
+
+    @ensure_cv2_image_for_annotation
+    def annotate(self, scene: ImageType, detections: Detections) -> ImageType:
+        """
+        Applies a colored overlay to the scene outside of the detected regions.
+
+        Args:
+            scene (ImageType): The image where masks will be drawn.
+                `ImageType` is a flexible type, accepting either `numpy.ndarray`
+                or `PIL.Image.Image`.
+            detections (Detections): Object detections to annotate.
+
+        Returns:
+            The annotated image, matching the type of `scene` (`numpy.ndarray`
+                or `PIL.Image.Image`)
+
+        Example:
+            ```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
+            )
+            ```
+
+        ![background-overlay-annotator-example](https://media.roboflow.com/
+        supervision-annotator-examples/background-color-annotator-example-purple.png)
+        """
+        assert isinstance(scene, np.ndarray)
+        colored_mask = np.full_like(scene, self.color.as_bgr(), dtype=np.uint8)
+
+        cv2.addWeighted(
+            scene, 1 - self.opacity, colored_mask, self.opacity, 0, dst=colored_mask
+        )
+
+        if detections.mask is None or self.force_box:
+            for x1, y1, x2, y2 in detections.xyxy.astype(int):
+                colored_mask[y1:y2, x1:x2] = scene[y1:y2, x1:x2]
+        else:
+            for mask in detections.mask:
+                colored_mask[mask] = scene[mask]
+
+        np.copyto(scene, colored_mask)
+        return scene

supervision/dataset/core.py

@@ -43,7 +43,10 @@ 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
 
@@ -174,18 +177,21 @@ class DetectionDataset(BaseDataset):
         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
@@ -396,7 +402,7 @@ 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:
@@ -455,10 +461,10 @@ 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, optional): If True, loads the annotations in OBB format.
+            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.
 
@@ -565,7 +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.
 
@@ -777,18 +783,21 @@ class ClassificationDataset(BaseDataset):
         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

supervision/dataset/formats/pascal_voc.py

@@ -147,7 +147,7 @@ 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:

supervision/dataset/formats/yolo.py

@@ -138,9 +138,9 @@ 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, optional): If True, loads the annotations in OBB format.
+        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.
 

supervision/detection/core.py

@@ -18,6 +18,11 @@ from supervision.detection.overlap_filter import (
     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,
@@ -244,10 +249,6 @@ 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 hasattr(ultralytics_results, "obb") and ultralytics_results.obb is not None:
@@ -267,6 +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(
@@ -438,15 +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.
@@ -477,36 +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
@@ -583,10 +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)
@@ -1157,10 +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.
 
@@ -1212,9 +1213,9 @@ class Detections:
         Perform non-maximum merging on the current set of object detections.
 
         Args:
-            threshold (float, optional): The intersection-over-union threshold
+            threshold (float): The intersection-over-union threshold
                 to use for non-maximum merging. Defaults to 0.5.
-            class_agnostic (bool, optional): Whether to perform class-agnostic
+            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.
 

supervision/detection/overlap_filter.py

@@ -1,3 +1,5 @@
+from __future__ import annotations
+
 from enum import Enum
 from typing import List, Union
 
@@ -53,9 +55,9 @@ def mask_non_max_suppression(
         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
+        iou_threshold (float): The intersection-over-union threshold
             to use for non-maximum suppression.
-        mask_dimension (int, optional): The dimension to which the masks should be
+        mask_dimension (int): The dimension to which the masks should be
             resized before computing IOU values. Defaults to 640.
 
     Returns:
@@ -101,7 +103,7 @@ def box_non_max_suppression(
         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
+        iou_threshold (float): The intersection-over-union threshold
             to use for non-maximum suppression.
 
     Returns:
@@ -156,7 +158,7 @@ def group_overlapping_boxes(
         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, optional): The intersection-over-union threshold
+        iou_threshold (float): The intersection-over-union threshold
             to use for non-maximum suppression. Defaults to 0.5.
 
     Returns:
@@ -200,7 +202,7 @@ def box_non_max_merge(
             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, optional): The intersection-over-union threshold
+        iou_threshold (float): The intersection-over-union threshold
             to use for non-maximum suppression. Defaults to 0.5.
 
     Returns:
@@ -248,16 +250,21 @@ class OverlapFilter(Enum):
     NON_MAX_SUPPRESSION = "non_max_suppression"
     NON_MAX_MERGE = "non_max_merge"
 
+    @classmethod
+    def list(cls):
+        return list(map(lambda c: c.value, cls))
 
-def validate_overlap_filter(
-    strategy: Union[OverlapFilter, str],
-) -> OverlapFilter:
-    if isinstance(strategy, str):
-        try:
-            strategy = OverlapFilter(strategy.lower())
-        except ValueError:
-            raise ValueError(
-                f"Invalid strategy value: {strategy}. Must be one of "
-                f"{[e.value for e in OverlapFilter]}"
-            )
-    return strategy
+    @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

@@ -4,11 +4,16 @@ 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.overlap_filter import OverlapFilter, validate_overlap_filter
-from supervision.detection.utils import move_boxes, move_masks
+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
+from supervision.utils.internal import (
+    SupervisionWarnings,
+    deprecated_parameter,
+    warn_deprecated,
+)
 
 
 def move_detections(
@@ -28,6 +33,10 @@ def move_detections(
         (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(
@@ -49,11 +58,16 @@ 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)`.
-        overlap_filter_strategy (Union[OverlapFilter, str]): Strategy for
+        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.
@@ -68,23 +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),
-        overlap_filter_strategy: Union[
-            OverlapFilter, str
-        ] = OverlapFilter.NON_MAX_SUPPRESSION,
+        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,
     ):
-        overlap_filter_strategy = validate_overlap_filter(overlap_filter_strategy)
+        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.overlap_ratio_wh = overlap_ratio_wh
         self.iou_threshold = iou_threshold
-        self.overlap_filter_strategy = overlap_filter_strategy
+        self.overlap_filter = OverlapFilter.from_value(overlap_filter)
         self.callback = callback
         self.thread_workers = thread_workers
 
@@ -129,6 +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:
@@ -139,15 +168,15 @@ class InferenceSlicer:
                 detections_list.append(future.result())
 
         merged = Detections.merge(detections_list=detections_list)
-        if self.overlap_filter_strategy == OverlapFilter.NONE:
+        if self.overlap_filter == OverlapFilter.NONE:
             return merged
-        elif self.overlap_filter_strategy == OverlapFilter.NON_MAX_SUPPRESSION:
+        elif self.overlap_filter == OverlapFilter.NON_MAX_SUPPRESSION:
             return merged.with_nms(threshold=self.iou_threshold)
-        elif self.overlap_filter_strategy == OverlapFilter.NON_MAX_MERGE:
+        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_strategy}",
+                f"Invalid overlap filter strategy: {self.overlap_filter}",
                 category=SupervisionWarnings,
             )
             return merged
@@ -177,7 +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,
@@ -186,12 +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
@@ -206,10 +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)
@@ -221,3 +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],
@@ -147,7 +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/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

@@ -106,7 +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`.
@@ -596,6 +596,61 @@ def move_boxes(
     return xyxy + np.hstack([offset, offset])
 
 
+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],

supervision/draw/color.py

@@ -352,11 +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

@@ -142,7 +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.
@@ -172,14 +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:

supervision/keypoint/annotators.py

@@ -34,8 +34,8 @@ 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
@@ -78,6 +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
 
@@ -108,8 +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.
         """
@@ -155,6 +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
 
@@ -194,7 +196,7 @@ class VertexLabelAnnotator:
     def __init__(
         self,
         color: Union[Color, List[Color]] = Color.ROBOFLOW,
-        text_color: Color = Color.WHITE,
+        text_color: Union[Color, List[Color]] = Color.WHITE,
         text_scale: float = 0.5,
         text_thickness: int = 1,
         text_padding: int = 10,
@@ -202,25 +204,30 @@ class VertexLabelAnnotator:
     ):
         """
         Args:
-            color (Union[Color, List[Color]], optional): The color to use for each
+            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 (Color, optional): The color to use for the labels.
-            text_scale (float, optional): The scale of the text.
-            text_thickness (int, optional): The thickness of the text.
-            text_padding (int, optional): The padding around the text.
-            border_radius (int, optional): The radius of the rounded corners of the
+            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: Color = text_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: List[str] = None
+        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
@@ -232,7 +239,7 @@ class VertexLabelAnnotator:
                 `PIL.Image.Image`.
             key_points (KeyPoints): A collection of key points where each key point
                 consists of x and y coordinates.
-            labels (List[str], optional): A list of labels to be displayed on the
+            labels (Optional[List[str]]): A list of labels to be displayed on the
                 annotated image. If not provided, keypoint indices will be used.
 
         Returns:
@@ -303,6 +310,7 @@ class VertexLabelAnnotator:
         ![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
@@ -321,12 +329,19 @@ class VertexLabelAnnotator:
             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(
@@ -344,7 +359,9 @@ class VertexLabelAnnotator:
 
         xyxy_padded = pad_boxes(xyxy=xyxy, px=self.text_padding)
 
-        for text, color, box, box_padded in zip(labels, colors, xyxy, xyxy_padded):
+        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),
@@ -354,10 +371,10 @@ class VertexLabelAnnotator:
             cv2.putText(
                 img=scene,
                 text=text,
-                org=(box[0], box[1] + self.text_padding),
+                org=(box[0], box[3]),
                 fontFace=font,
                 fontScale=self.text_scale,
-                color=self.text_color.as_rgb(),
+                color=text_color.as_bgr(),
                 thickness=self.text_thickness,
                 lineType=cv2.LINE_AA,
             )
@@ -389,7 +406,7 @@ class VertexLabelAnnotator:
     @staticmethod
     def preprocess_and_validate_labels(
         labels: Optional[List[str]], points_count: int, skeletons_count: int
-    ) -> np.array:
+    ) -> np.ndarray:
         if labels and len(labels) != points_count:
             raise ValueError(
                 f"Number of labels ({len(labels)}) must match number of key points "
@@ -405,7 +422,7 @@ class VertexLabelAnnotator:
         colors: Optional[Union[Color, List[Color]]],
         points_count: int,
         skeletons_count: int,
-    ) -> np.array:
+    ) -> 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 "

supervision/keypoint/core.py

@@ -88,15 +88,17 @@ class KeyPoints:
         ```
 
     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]
@@ -132,7 +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 (

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

@@ -7,7 +7,6 @@ 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):
@@ -197,48 +196,24 @@ 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.
-        minimum_consecutive_frames (int, optional): Number of consecutive frames that an object must
+        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,

supervision/utils/image.py

@@ -158,7 +158,7 @@ def resize_image(
             accepting either `numpy.ndarray` or `PIL.Image.Image`.
         resolution_wh (Tuple[int, int]): The target resolution as
             `(width, height)`.
-        keep_aspect_ratio (bool, optional): Flag to maintain the image's original
+        keep_aspect_ratio (bool): Flag to maintain the image's original
             aspect ratio. Defaults to `False`.
 
     Returns:
@@ -270,6 +270,7 @@ def letterbox_image(
 
     ![letterbox_image](https://media.roboflow.com/supervision-docs/letterbox-image.png){ align=center width="800" }
     """  # noqa E501 // docs
+    assert isinstance(image, np.ndarray)
     color = unify_to_bgr(color=color)
     resized_image = resize_image(
         image=image, resolution_wh=resolution_wh, keep_aspect_ratio=True
@@ -279,7 +280,7 @@ def letterbox_image(
     padding_bottom = resolution_wh[1] - height_new - padding_top
     padding_left = (resolution_wh[0] - width_new) // 2
     padding_right = resolution_wh[0] - width_new - padding_left
-    return cv2.copyMakeBorder(
+    image_with_borders = cv2.copyMakeBorder(
         resized_image,
         padding_top,
         padding_bottom,
@@ -289,6 +290,14 @@ def letterbox_image(
         value=color,
     )
 
+    if image.shape[2] == 4:
+        image[:padding_top, :, 3] = 0
+        image[height_new - padding_bottom :, :, 3] = 0
+        image[:, :padding_left, 3] = 0
+        image[:, width_new - padding_right :, 3] = 0
+
+    return image_with_borders
+
 
 def overlay_image(
     image: npt.NDArray[np.uint8],
@@ -341,9 +350,20 @@ def overlay_image(
     crop_x_max = image_width - max((anchor_x + image_width) - scene_width, 0)
     crop_y_max = image_height - max((anchor_y + image_height) - scene_height, 0)
 
-    image[y_min:y_max, x_min:x_max] = overlay[
-        crop_y_min:crop_y_max, crop_x_min:crop_x_max
-    ]
+    if overlay.shape[2] == 4:
+        b, g, r, alpha = cv2.split(
+            overlay[crop_y_min:crop_y_max, crop_x_min:crop_x_max]
+        )
+        alpha = alpha[:, :, None] / 255.0
+        overlay_color = cv2.merge((b, g, r))
+
+        roi = image[y_min:y_max, x_min:x_max]
+        roi[:] = roi * (1 - alpha) + overlay_color * alpha
+        image[y_min:y_max, x_min:x_max] = roi
+    else:
+        image[y_min:y_max, x_min:x_max] = overlay[
+            crop_y_min:crop_y_max, crop_x_min:crop_x_max
+        ]
 
     return image
 
@@ -360,9 +380,9 @@ class ImageSink:
 
         Args:
             target_dir_path (str): The target directory where images will be saved.
-            overwrite (bool, optional): Whether to overwrite the existing directory.
+            overwrite (bool): Whether to overwrite the existing directory.
                 Defaults to False.
-            image_name_pattern (str, optional): The image file name pattern.
+            image_name_pattern (str): The image file name pattern.
                 Defaults to "image_{:05d}.png".
 
         Examples:
@@ -399,7 +419,7 @@ class ImageSink:
         Args:
             image (np.ndarray): The image to be saved. The image must be in BGR color
                 format.
-            image_name (str, optional): The name to use for the saved image.
+            image_name (Optional[str]): The name to use for the saved image.
                 If not provided, a name will be
                 generated using the `image_name_pattern`.
         """

supervision/utils/internal.py

@@ -2,7 +2,7 @@ import functools
 import inspect
 import os
 import warnings
-from typing import Any, Callable, Set
+from typing import Any, Callable, Generic, Optional, Set, TypeVar
 
 
 class SupervisionWarnings(Warning):
@@ -56,9 +56,9 @@ def deprecated_parameter(
     Parameters:
         old_parameter (str): The name of the deprecated parameter.
         new_parameter (str): The name of the parameter that should be used instead.
-        map_function (Callable, optional): A function used to map the value of the old
+        map_function (Callable): A function used to map the value of the old
             parameter to the new parameter. Defaults to the identity function.
-        warning_message (str, optional): The warning message to be displayed when the
+        warning_message (str): The warning message to be displayed when the
             deprecated parameter is used. Defaults to a generic warning message with
             placeholders for the old parameter, new parameter, and function name.
         **message_kwargs: Additional keyword arguments that can be used to customize
@@ -122,7 +122,10 @@ def deprecated(reason: str):
     return decorator
 
 
-class classproperty(property):
+T = TypeVar("T")
+
+
+class classproperty(Generic[T]):
     """
     A decorator that combines @classmethod and @property.
     It allows a method to be accessed as a property of the class,
@@ -134,17 +137,27 @@ class classproperty(property):
             ...
     """
 
-    def __get__(self, owner_self: object, owner_cls: type) -> object:
+    def __init__(self, fget: Callable[..., T]):
+        """
+        Args:
+            The function that is called when the property is accessed.
+        """
+        self.fget = fget
+
+    def __get__(self, owner_self: Any, owner_cls: Optional[type] = None) -> T:
         """
         Override the __get__ method to return the result of the function call.
 
         Args:
-        owner_self: The instance through which the attribute was accessed, or None.
-        owner_cls: The class through which the attribute was accessed.
+            owner_self: The instance through which the attribute was accessed, or None.
+                Irrelevant for class properties.
+            owner_cls: The class through which the attribute was accessed.
 
         Returns:
-        The result of calling the function stored in 'fget' with 'owner_cls'.
+            The result of calling the function stored in 'fget' with 'owner_cls'.
         """
+        if self.fget is None:
+            raise AttributeError("unreadable attribute")
         return self.fget(owner_cls)
 
 

supervision/utils/notebook.py

@@ -17,7 +17,7 @@ def plot_image(
     Args:
         image (ImageType): The frame to be displayed ImageType
              is a flexible type, accepting either `numpy.ndarray` or `PIL.Image.Image`.
-        size (Tuple[int, int]): The size of the plot.
+        size (Tuple[int, int]): The size of the plot in inches.
         cmap (str): the colormap to use for single channel images.
 
     Examples:

supervision/utils/video.py

@@ -19,7 +19,7 @@ class VideoInfo:
         width (int): width of the video in pixels
         height (int): height of the video in pixels
         fps (int): frames per second of the video
-        total_frames (int, optional): total number of frames in the video,
+        total_frames (Optional[int]): total number of frames in the video,
             default is None
 
     Examples:
@@ -116,7 +116,9 @@ class VideoSink:
         self.__writer.release()
 
 
-def _validate_and_setup_video(source_path: str, start: int, end: Optional[int]):
+def _validate_and_setup_video(
+    source_path: str, start: int, end: Optional[int], iterative_seek: bool = False
+):
     video = cv2.VideoCapture(source_path)
     if not video.isOpened():
         raise Exception(f"Could not open video at {source_path}")
@@ -125,12 +127,25 @@ def _validate_and_setup_video(source_path: str, start: int, end: Optional[int]):
         raise Exception("Requested frames are outbound")
     start = max(start, 0)
     end = min(end, total_frames) if end is not None else total_frames
-    video.set(cv2.CAP_PROP_POS_FRAMES, start)
+
+    if iterative_seek:
+        while start > 0:
+            success = video.grab()
+            if not success:
+                break
+            start -= 1
+    elif start > 0:
+        video.set(cv2.CAP_PROP_POS_FRAMES, start)
+
     return video, start, end
 
 
 def get_video_frames_generator(
-    source_path: str, stride: int = 1, start: int = 0, end: Optional[int] = None
+    source_path: str,
+    stride: int = 1,
+    start: int = 0,
+    end: Optional[int] = None,
+    iterative_seek: bool = False,
 ) -> Generator[np.ndarray, None, None]:
     """
     Get a generator that yields the frames of the video.
@@ -143,6 +158,9 @@ def get_video_frames_generator(
             video should generate frames
         end (Optional[int]): Indicates the ending position at which video
             should stop generating frames. If None, video will be read to the end.
+        iterative_seek (bool): If True, the generator will seek to the
+            `start` frame by grabbing each frame, which is much slower. This is a
+            workaround for videos that don't open at all when you set the `start` value.
 
     Returns:
         (Generator[np.ndarray, None, None]): A generator that yields the
@@ -156,7 +174,9 @@ def get_video_frames_generator(
             ...
         ```
     """
-    video, start, end = _validate_and_setup_video(source_path, start, end)
+    video, start, end = _validate_and_setup_video(
+        source_path, start, end, iterative_seek
+    )
     frame_position = start
     while True:
         success, frame = video.read()

test/annotators/test_utils.py

@@ -1,5 +1,4 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 from typing import Optional
 
 import numpy as np
@@ -7,6 +6,7 @@ import pytest
 
 from supervision.annotators.utils import ColorLookup, resolve_color_idx
 from supervision.detection.core import Detections
+from test.test_utils import mock_detections
 
 
 @pytest.mark.parametrize(

test/dataset/formats/test_pascal_voc.py

@@ -1,16 +1,16 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 from typing import List, Optional
 
-import defusedxml.ElementTree as ET
 import numpy as np
 import pytest
+from defusedxml import ElementTree
 
 from supervision.dataset.formats.pascal_voc import (
     detections_from_xml_obj,
     object_to_pascal_voc,
     parse_polygon_points,
 )
+from test.test_utils import mock_detections
 
 
 def are_xml_elements_equal(elem1, elem2):
@@ -36,7 +36,7 @@ def are_xml_elements_equal(elem1, elem2):
             np.array([0, 0, 10, 10]),
             "test",
             None,
-            ET.fromstring(
+            ElementTree.fromstring(
                 """<object><name>test</name><bndbox><xmin>1</xmin><ymin>1</ymin>
                 <xmax>11</xmax><ymax>11</ymax></bndbox></object>"""
             ),
@@ -46,7 +46,7 @@ def are_xml_elements_equal(elem1, elem2):
             np.array([0, 0, 10, 10]),
             "test",
             np.array([[0, 0], [10, 0], [10, 10], [0, 10]]),
-            ET.fromstring(
+            ElementTree.fromstring(
                 """<object><name>test</name><bndbox><xmin>1</xmin><ymin>1</ymin>
                 <xmax>11</xmax><ymax>11</ymax>
                 </bndbox><polygon><x1>1</x1><y1>1</y1><x2>11</x2>
@@ -73,7 +73,7 @@ def test_object_to_pascal_voc(
     "polygon_element, expected_result, exception",
     [
         (
-            ET.fromstring(
+            ElementTree.fromstring(
                 """<polygon><x1>0</x1><y1>0</y1><x2>10</x2><y2>0</y2><x3>10</x3>
                     <y3>10</y3><x4>0</x4><y4>10</y4></polygon>"""
             ),
@@ -160,6 +160,6 @@ def test_detections_from_xml_obj(
     xml_string, classes, resolution_wh, force_masks, expected_result, exception
 ):
     with exception:
-        root = ET.fromstring(xml_string)
+        root = ElementTree.fromstring(xml_string)
         result, _ = detections_from_xml_obj(root, classes, resolution_wh, force_masks)
         assert result == expected_result

test/dataset/formats/test_yolo.py

@@ -13,7 +13,7 @@ from supervision.dataset.formats.yolo import (
 from supervision.detection.core import Detections
 
 
-def _mock_simple_mask(resolution_wh: Tuple[int, int], box: List[int]) -> np.array:
+def _mock_simple_mask(resolution_wh: Tuple[int, int], box: List[int]) -> np.ndarray:
     x_min, y_min, x_max, y_max = box
     mask = np.full(resolution_wh, False, dtype=bool)
     mask[y_min:y_max, x_min:x_max] = True

test/dataset/test_core.py

@@ -1,11 +1,10 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 from typing import List, Optional
 
-import numpy as np
 import pytest
 
 from supervision import DetectionDataset
+from test.test_utils import mock_detections
 
 
 @pytest.mark.parametrize(
@@ -13,29 +12,29 @@ from supervision import DetectionDataset
     [
         (
             [],
-            DetectionDataset(classes=[], images={}, annotations={}),
+            DetectionDataset(classes=[], images=[], annotations={}),
             DoesNotRaise(),
         ),  # empty dataset list
         (
-            [DetectionDataset(classes=[], images={}, annotations={})],
-            DetectionDataset(classes=[], images={}, annotations={}),
+            [DetectionDataset(classes=[], images=[], annotations={})],
+            DetectionDataset(classes=[], images=[], annotations={}),
             DoesNotRaise(),
         ),  # single empty dataset
         (
             [
-                DetectionDataset(classes=["dog", "person"], images={}, annotations={}),
-                DetectionDataset(classes=["dog", "person"], images={}, annotations={}),
+                DetectionDataset(classes=["dog", "person"], images=[], annotations={}),
+                DetectionDataset(classes=["dog", "person"], images=[], annotations={}),
             ],
-            DetectionDataset(classes=["dog", "person"], images={}, annotations={}),
+            DetectionDataset(classes=["dog", "person"], images=[], annotations={}),
             DoesNotRaise(),
         ),  # two datasets; no images and annotations, the same classes
         (
             [
-                DetectionDataset(classes=["dog", "person"], images={}, annotations={}),
-                DetectionDataset(classes=["cat"], images={}, annotations={}),
+                DetectionDataset(classes=["dog", "person"], images=[], annotations={}),
+                DetectionDataset(classes=["cat"], images=[], annotations={}),
             ],
             DetectionDataset(
-                classes=["cat", "dog", "person"], images={}, annotations={}
+                classes=["cat", "dog", "person"], images=[], annotations={}
             ),
             DoesNotRaise(),
         ),  # two datasets; no images and annotations, different classes
@@ -43,10 +42,7 @@ from supervision import DetectionDataset
             [
                 DetectionDataset(
                     classes=["dog", "person"],
-                    images={
-                        "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                        "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    },
+                    images=["image-1.png", "image-2.png"],
                     annotations={
                         "image-1.png": mock_detections(
                             xyxy=[[0, 0, 10, 10]], class_id=[0]
@@ -56,14 +52,11 @@ from supervision import DetectionDataset
                         ),
                     },
                 ),
-                DetectionDataset(classes=[], images={}, annotations={}),
+                DetectionDataset(classes=[], images=[], annotations={}),
             ],
             DetectionDataset(
                 classes=["dog", "person"],
-                images={
-                    "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                },
+                images=["image-1.png", "image-2.png"],
                 annotations={
                     "image-1.png": mock_detections(xyxy=[[0, 0, 10, 10]], class_id=[0]),
                     "image-2.png": mock_detections(xyxy=[[0, 0, 10, 10]], class_id=[1]),
@@ -75,10 +68,7 @@ from supervision import DetectionDataset
             [
                 DetectionDataset(
                     classes=["dog", "person"],
-                    images={
-                        "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                        "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    },
+                    images=["image-1.png", "image-2.png"],
                     annotations={
                         "image-1.png": mock_detections(
                             xyxy=[[0, 0, 10, 10]], class_id=[0]
@@ -88,14 +78,11 @@ from supervision import DetectionDataset
                         ),
                     },
                 ),
-                DetectionDataset(classes=["cat"], images={}, annotations={}),
+                DetectionDataset(classes=["cat"], images=[], annotations={}),
             ],
             DetectionDataset(
                 classes=["cat", "dog", "person"],
-                images={
-                    "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                },
+                images=["image-1.png", "image-2.png"],
                 annotations={
                     "image-1.png": mock_detections(xyxy=[[0, 0, 10, 10]], class_id=[1]),
                     "image-2.png": mock_detections(xyxy=[[0, 0, 10, 10]], class_id=[2]),
@@ -107,10 +94,7 @@ from supervision import DetectionDataset
             [
                 DetectionDataset(
                     classes=["dog", "person"],
-                    images={
-                        "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                        "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    },
+                    images=["image-1.png", "image-2.png"],
                     annotations={
                         "image-1.png": mock_detections(
                             xyxy=[[0, 0, 10, 10]], class_id=[0]
@@ -122,9 +106,7 @@ from supervision import DetectionDataset
                 ),
                 DetectionDataset(
                     classes=["cat"],
-                    images={
-                        "image-3.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    },
+                    images=["image-3.png"],
                     annotations={
                         "image-3.png": mock_detections(
                             xyxy=[[0, 0, 10, 10]], class_id=[0]
@@ -134,11 +116,7 @@ from supervision import DetectionDataset
             ],
             DetectionDataset(
                 classes=["cat", "dog", "person"],
-                images={
-                    "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    "image-3.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                },
+                images=["image-1.png", "image-2.png", "image-3.png"],
                 annotations={
                     "image-1.png": mock_detections(xyxy=[[0, 0, 10, 10]], class_id=[1]),
                     "image-2.png": mock_detections(xyxy=[[0, 0, 10, 10]], class_id=[2]),
@@ -151,10 +129,7 @@ from supervision import DetectionDataset
             [
                 DetectionDataset(
                     classes=["dog", "person"],
-                    images={
-                        "image-1.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                        "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    },
+                    images=["image-1.png", "image-2.png"],
                     annotations={
                         "image-1.png": mock_detections(
                             xyxy=[[0, 0, 10, 10]], class_id=[0]
@@ -166,10 +141,7 @@ from supervision import DetectionDataset
                 ),
                 DetectionDataset(
                     classes=["dog", "person"],
-                    images={
-                        "image-2.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                        "image-3.png": np.zeros((100, 100, 3), dtype=np.uint8),
-                    },
+                    images=["image-2.png", "image-3.png"],
                     annotations={
                         "image-2.png": mock_detections(
                             xyxy=[[0, 0, 10, 10]], class_id=[0]

test/dataset/test_utils.py

@@ -1,5 +1,4 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 from typing import Dict, List, Optional, Tuple, TypeVar
 
 import numpy as np
@@ -15,6 +14,7 @@ from supervision.dataset.utils import (
     rle_to_mask,
     train_test_split,
 )
+from test.test_utils import mock_detections
 
 T = TypeVar("T")
 
@@ -278,7 +278,7 @@ def test_map_detections_class_id(
             np.array([[[]]]).astype(bool),
             None,
             pytest.raises(AssertionError),
-        ),  # raises AssertionError because mask dimentionality is not 2D
+        ),  # raises AssertionError because mask dimensionality is not 2D
         (
             np.array([[]]).astype(bool),
             None,

test/detection/test_core.py

@@ -1,5 +1,4 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 from typing import List, Optional, Union
 
 import numpy as np
@@ -7,6 +6,7 @@ import pytest
 
 from supervision.detection.core import Detections, merge_inner_detection_object_pair
 from supervision.geometry.core import Position
+from test.test_utils import mock_detections
 
 PREDICTIONS = np.array(
     [

test/detection/test_csv.py

@@ -1,11 +1,11 @@
 import csv
 import os
-from test.test_utils import mock_detections
 from typing import Any, Dict, List
 
 import pytest
 
 import supervision as sv
+from test.test_utils import mock_detections
 
 
 @pytest.mark.parametrize(

test/detection/test_json.py

@@ -1,11 +1,11 @@
 import json
 import os
-from test.test_utils import mock_detections
 from typing import Any, Dict, List
 
 import pytest
 
 import supervision as sv
+from test.test_utils import mock_detections
 
 
 @pytest.mark.parametrize(

test/detection/test_line_counter.py

@@ -1,11 +1,11 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 from typing import List, Optional, Tuple
 
 import pytest
 
 from supervision import LineZone
 from supervision.geometry.core import Point, Position, Vector
+from test.test_utils import mock_detections
 
 
 @pytest.mark.parametrize(

test/detection/test_polygonzone.py

@@ -1,10 +1,10 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import mock_detections
 
 import numpy as np
 import pytest
 
 import supervision as sv
+from test.test_utils import mock_detections
 
 DETECTION_BOXES = np.array(
     [
@@ -26,7 +26,6 @@ DETECTIONS = mock_detections(
 )
 
 POLYGON = np.array([[100, 100], [200, 100], [200, 200], [100, 200]])
-FRAME_RESOLUTION = (300, 300)
 
 
 @pytest.mark.parametrize(
@@ -36,7 +35,6 @@ FRAME_RESOLUTION = (300, 300)
             DETECTIONS,
             sv.PolygonZone(
                 POLYGON,
-                FRAME_RESOLUTION,
                 triggering_anchors=(
                     sv.Position.TOP_LEFT,
                     sv.Position.TOP_RIGHT,
@@ -53,7 +51,6 @@ FRAME_RESOLUTION = (300, 300)
             DETECTIONS,
             sv.PolygonZone(
                 POLYGON,
-                FRAME_RESOLUTION,
             ),
             np.array(
                 [False, False, True, True, True, True, False, False, False], dtype=bool
@@ -64,7 +61,6 @@ FRAME_RESOLUTION = (300, 300)
             DETECTIONS,
             sv.PolygonZone(
                 POLYGON,
-                FRAME_RESOLUTION,
                 triggering_anchors=[sv.Position.BOTTOM_CENTER],
             ),
             np.array(
@@ -76,7 +72,6 @@ FRAME_RESOLUTION = (300, 300)
             sv.Detections.empty(),
             sv.PolygonZone(
                 POLYGON,
-                FRAME_RESOLUTION,
             ),
             np.array([], dtype=bool),
             DoesNotRaise(),
@@ -107,4 +102,4 @@ def test_polygon_zone_trigger(
 )
 def test_polygon_zone_initialization(polygon, triggering_anchors, exception):
     with exception:
-        sv.PolygonZone(polygon, FRAME_RESOLUTION, triggering_anchors=triggering_anchors)
+        sv.PolygonZone(polygon, triggering_anchors=triggering_anchors)

test/detection/test_utils.py

@@ -975,7 +975,7 @@ def test_get_data_item(
             ),
             False,
             DoesNotRaise(),
-        ),  # foreground object in 2 seperate elements
+        ),  # foreground object in 2 separate elements
         (
             np.array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]).astype(
                 bool
@@ -1032,7 +1032,7 @@ def test_contains_holes(
             4,
             True,
             DoesNotRaise(),
-        ),  # foreground object in 2 seperate elements
+        ),  # foreground object in 2 separate elements
         (
             np.array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]).astype(
                 bool

test/metrics/test_detection.py

@@ -1,5 +1,4 @@
 from contextlib import ExitStack as DoesNotRaise
-from test.test_utils import assert_almost_equal, mock_detections
 from typing import Optional, Union
 
 import numpy as np
@@ -11,6 +10,7 @@ from supervision.metrics.detection import (
     MeanAveragePrecision,
     detections_to_tensor,
 )
+from test.test_utils import assert_almost_equal, mock_detections
 
 CLASSES = np.arange(80)
 NUM_CLASSES = len(CLASSES)

test/test_utils.py

@@ -1,13 +1,14 @@
 from typing import Any, Dict, List, Optional
 
 import numpy as np
+import numpy.typing as npt
 
 from supervision.detection.core import Detections
 from supervision.keypoint.core import KeyPoints
 
 
 def mock_detections(
-    xyxy: List[List[float]],
+    xyxy: npt.NDArray[np.float32],
     mask: Optional[List[np.ndarray]] = None,
     confidence: Optional[List[float]] = None,
     class_id: Optional[List[int]] = None,
@@ -32,7 +33,7 @@ def mock_detections(
 
 
 def mock_keypoints(
-    xy: List[List[float]],
+    xy: npt.NDArray[np.float32],
     confidence: Optional[List[float]] = None,
     class_id: Optional[List[int]] = None,
     data: Optional[Dict[str, List[Any]]] = None,