Merge pull request #1790 from roboflow/feature/support_for_qwen_2_5_vl

support for converting Qwen2.5-VL into sv.Detections

.github/workflows/poetry-test.yml

@@ -1,42 +0,0 @@
-name: 🔧 Poetry Check and Installation Test Workflow
-on:
-  push:
-    paths:
-      - "poetry.lock"
-      - "pyproject.toml"
-  pull_request:
-    paths:
-      - "poetry.lock"
-      - "pyproject.toml"
-  workflow_dispatch:
-
-jobs:
-  poetry-tests:
-    name: Poetry install and check
-    timeout-minutes: 10
-    strategy:
-      fail-fast: false
-      matrix:
-        os: [ubuntu-latest, windows-latest, macos-latest]
-        python-version: ["3.8", "3.9", "3.10", "3.11", "3.12", "3.13"]
-    runs-on: ${{ matrix.os }}
-    steps:
-      - name: 📥 Checkout the repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: 📜 Setup Poetry
-        uses: abatilo/actions-poetry@e78f54a89cb052fff327414dd9ff010b5d2b4dbd # v3.0.1
-
-      - name: 🔍 Check the correctness of the project config
-        run: poetry check
-
-      - name: 🚀 Do Install the package Test
-        run: poetry install --with dev --extras metrics
-
-      - name: 🧪 Run the Import test
-        run: poetry run python -c "import supervision; from supervision import assets; from supervision import metrics; print(supervision.__version__)"

.github/workflows/publish-docs.yml

@@ -33,17 +33,25 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+      - name: 🐍 Install uv and set Python ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@4db96194c378173c656ce18a155ffc14a9fc4355 # v5.2.2
         with:
           python-version: ${{ matrix.python-version }}
 
-      - name: 📜 Setup Poetry
-        uses: abatilo/actions-poetry@e78f54a89cb052fff327414dd9ff010b5d2b4dbd # v3.0.1
+      - name: 🔑 Create GitHub App token (mkdocs)
+        id: mkdocs_token
+        uses: actions/create-github-app-token@c1a285145b9d317df6ced56c09f525b5c2b6f755 # v1.11.1
+        with:
+          app-id: ${{ secrets.MKDOCS_APP_ID }}
+          private-key: ${{ secrets.MKDOCS_PEM }}
+          owner: roboflow
+          repositories: mkdocs-material-insiders
 
-      - name: 📦 Install dependencies
+      - name: 🏗️ Install dependencies
         run: |
-          poetry install --with=docs
+          uv pip install -r pyproject.toml --extra docs
+          # Install mkdocs-material-insiders using the GitHub App token
+          uv pip install "git+https://roboflow:${{ steps.mkdocs_token.outputs.token }}@github.com/roboflow/mkdocs-material-insiders.git@9.5.49-insiders-4.53.14#egg=mkdocs-material[imaging]"
 
       - name: ⚙️ Configure git for github-actions
         run: |
@@ -53,10 +61,10 @@ jobs:
       - name: 🚀 Deploy Development Docs
         if: (github.event_name == 'push' && github.ref == 'refs/heads/develop') || github.event_name == 'workflow_dispatch'
         run: |
-          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} poetry run mike deploy --push develop
+          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} uv run mike deploy --push develop
 
       - name: 🚀 Deploy Release Docs
         if: github.event_name == 'release' && github.event.action == 'published'
         run: |
           latest_tag=$(git describe --tags `git rev-list --tags --max-count=1`)
-          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} poetry run mike deploy --push --update-aliases $latest_tag latest
+          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} uv run mike deploy --push --update-aliases $latest_tag latest

.github/workflows/publish-pre-release.yml

@@ -0,0 +1,45 @@
+name: Publish Supervision Pre-Releases to PyPI
+
+on:
+  push:
+    tags:
+      - "[0-9]+.[0-9]+[0-9]+.[0-9]+a[0-9]"
+      - "[0-9]+.[0-9]+[0-9]+.[0-9]+b[0-9]"
+      - "[0-9]+.[0-9]+[0-9]+.[0-9]+rc[0-9]"
+  workflow_dispatch:
+
+permissions: {} # Explicitly remove all permissions by default
+
+jobs:
+  publish-pre-release:
+    name: Publish Pre-release Package
+    runs-on: ubuntu-latest
+    environment:
+      name: test
+      url: https://pypi.org/project/supervision/
+    timeout-minutes: 10
+    permissions:
+      id-token: write # Required for PyPI publishing
+      contents: read # Required for checkout
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+    steps:
+      - name: 📥 Checkout the repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+
+      - name: 🐍 Install uv and set Python version ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@4db96194c378173c656ce18a155ffc14a9fc4355 # v5.2.2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: 🏗️ Build source and wheel distributions
+        run: |
+          uv pip install -r pyproject.toml --extra build
+          uv build
+          uv run twine check --strict dist/*
+
+      - name: 🚀 Publish to PyPi
+        uses: pypa/gh-action-pypi-publish@76f52bc884231f62b9a034ebfe128415bbaabdfc # v1.12.4
+        with:
+          attestations: true

.github/workflows/publish-release.yml

@@ -1,4 +1,4 @@
-name: Publish Supervision Releases to PyPI and TestPyPI
+name: Publish Supervision Releases to PyPI
 
 on:
   push:
@@ -9,10 +9,12 @@ on:
 permissions: {} # Explicitly remove all permissions by default
 
 jobs:
-  build-and-publish-pre-release:
+  publish-release:
     name: Publish Release Package
     runs-on: ubuntu-latest
-    environment: release
+    environment:
+      name: release
+      url: https://pypi.org/project/supervision/
     timeout-minutes: 10
     permissions:
       id-token: write # Required for PyPI publishing
@@ -24,27 +26,18 @@ jobs:
       - name: 📥 Checkout the repository
         uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+      - name: 🐍 Install uv and set Python version ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@4db96194c378173c656ce18a155ffc14a9fc4355 # v5.2.2
         with:
           python-version: ${{ matrix.python-version }}
 
-      - name: 📜 Setup Poetry
-        uses: abatilo/actions-poetry@e78f54a89cb052fff327414dd9ff010b5d2b4dbd # v3.0.1
-
       - name: 🏗️ Build source and wheel distributions
         run: |
-          poetry install --with=build
-          poetry build
-          twine check --strict dist/*
+          uv pip install -r pyproject.toml --extra build
+          uv build
+          uv run twine check --strict dist/*
 
       - name: 🚀 Publish to PyPi
         uses: pypa/gh-action-pypi-publish@67339c736fd9354cd4f8cb0b744f2b82a74b5c70 # v1.12.3
         with:
           attestations: true
-
-      - name: 🚀 Publish to Test-PyPi
-        uses: pypa/gh-action-pypi-publish@67339c736fd9354cd4f8cb0b744f2b82a74b5c70 # v1.12.3
-        with:
-          repository-url: https://test.pypi.org/legacy/
-          attestations: true

.github/workflows/publish-test.yml

@@ -1,52 +0,0 @@
-name: Publish Supervision Pre-Releases to PyPI and TestPyPI
-
-on:
-  push:
-    tags:
-      - "[0-9]+.[0-9]+[0-9]+.[0-9]+a[0-9]"
-      - "[0-9]+.[0-9]+[0-9]+.[0-9]+b[0-9]"
-      - "[0-9]+.[0-9]+[0-9]+.[0-9]+rc[0-9]"
-  workflow_dispatch:
-
-permissions: {} # Explicitly remove all permissions by default
-
-jobs:
-  build-and-publish-pre-release:
-    name: Publish  Pre-releasePackage
-    runs-on: ubuntu-latest
-    environment: release
-    timeout-minutes: 10
-    permissions:
-      id-token: write # Required for PyPI publishing
-      contents: read # Required for checkout
-    strategy:
-      matrix:
-        python-version: ["3.10"]
-    steps:
-      - name: 📥 Checkout the repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: 📜 Setup Poetry
-        uses: abatilo/actions-poetry@e78f54a89cb052fff327414dd9ff010b5d2b4dbd # v3.0.1
-
-      - name: 🏗️ Build source and wheel distributions
-        run: |
-          poetry install --with=build
-          poetry build
-          twine check --strict dist/*
-
-      - name: 🚀 Publish to PyPi
-        uses: pypa/gh-action-pypi-publish@67339c736fd9354cd4f8cb0b744f2b82a74b5c70 # v1.12.3
-        with:
-          attestations: true
-
-      - name: 🚀 Publish to Test-PyPi
-        uses: pypa/gh-action-pypi-publish@67339c736fd9354cd4f8cb0b744f2b82a74b5c70 # v1.12.3
-        with:
-          repository-url: https://test.pypi.org/legacy/
-          attestations: true

.github/workflows/publish-testpypi.yml

@@ -0,0 +1,41 @@
+name: Publish Supervision Releases to TestPyPI
+
+on:
+  workflow_dispatch:
+
+permissions: {} # Explicitly remove all permissions by default
+
+jobs:
+  publish-testpypi:
+    name: Publish Release Package
+    runs-on: ubuntu-latest
+    environment:
+      name: release
+      url: https://pypi.org/project/supervision/
+    timeout-minutes: 10
+    permissions:
+      id-token: write # Required for PyPI publishing
+      contents: read # Required for checkout
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+    steps:
+      - name: 📥 Checkout the repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+
+      - name: 🐍 Install uv and set Python version ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@4db96194c378173c656ce18a155ffc14a9fc4355 # v5.2.2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: 🏗️ Build source and wheel distributions
+        run: |
+          uv pip install -r pyproject.toml --extra build
+          uv build
+          uv run twine check --strict dist/*
+
+      - name: 🚀 Publish to Test-PyPi
+        uses: pypa/gh-action-pypi-publish@76f52bc884231f62b9a034ebfe128415bbaabdfc # v1.12.4
+        with:
+          repository-url: https://test.pypi.org/legacy/
+          attestations: true

.github/workflows/test-doc.yml

@@ -23,18 +23,15 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
+      - name: 🐍 Install uv and set Python ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@4db96194c378173c656ce18a155ffc14a9fc4355 # v5.2.2
         with:
           python-version: ${{ matrix.python-version }}
 
-      - name: 📜 Setup Poetry
-        uses: abatilo/actions-poetry@e78f54a89cb052fff327414dd9ff010b5d2b4dbd # v3.0.1
-
       - name: 🏗️ Install dependencies
         run: |
-          poetry install --with=docs
+          uv pip install -r pyproject.toml --extra docs
+
 
       - name: 🧪 Test Docs Build
-        run: |
-          poetry run mkdocs build --verbose
+        run: uv run mkdocs build --verbose

.github/workflows/test.yml

@@ -1,47 +0,0 @@
-name: Test WorkFlow
-
-on:
-  pull_request:
-    branches: [main, develop]
-
-permissions:
-  contents: read
-  checks: write # Required for test reporting
-  pull-requests: write # Required for PR comments
-
-jobs:
-  pytest:
-    name: PyTest
-    timeout-minutes: 10
-    strategy:
-      matrix:
-        os: [ubuntu-latest, windows-latest, macos-latest]
-        python-version: ["3.8", "3.9", "3.10", "3.11", "3.12", "3.13"]
-      fail-fast: false # Continue with other versions if one fails
-    runs-on: ${{ matrix.os }}
-    steps:
-      - name: 📥 Checkout the repository
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: 📜 Setup Poetry
-        uses: abatilo/actions-poetry@e78f54a89cb052fff327414dd9ff010b5d2b4dbd # v3.0.1
-
-      - name: 📦 Install dependencies
-        run: |
-          poetry install --with=test
-
-      - name: Run pytest
-        uses: pavelzw/pytest-action@510c5e90c360a185039bea56ce8b3e7e51a16507 # v2.2.0
-        with:
-          verbose: true
-          emoji: true
-          job-summary: true
-          custom-arguments: "-q"
-          click-to-expand: true
-          report-title: "Test Report"
-          custom-pytest: "poetry run pytest"

.github/workflows/uv-test.yml

@@ -0,0 +1,33 @@
+name: 🔧 Pytest/Test Workflow
+
+on:
+  pull_request:
+    branches: [main, develop]
+
+jobs:
+  run-tests:
+    name: Import Test and Pytest Run
+    timeout-minutes: 10
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, windows-latest, macos-latest]
+        python-version: ["3.8", "3.9", "3.10", "3.11", "3.12", "3.13"]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - name: 📥 Checkout the repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+
+      - name: 🐍 Install uv and set Python version ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@4db96194c378173c656ce18a155ffc14a9fc4355 # v5.2.2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: 🚀 Install Packages
+        run: uv pip install -r pyproject.toml --extra dev --extra docs --extra metrics
+
+      - name: 🧪 Run the Import test
+        run: uv run python -c "import supervision; from supervision import assets; from supervision import metrics; print(supervision.__version__)"
+
+      - name: 🧪 Run the Test
+        run: uv run pytest

.pre-commit-config.yaml

@@ -32,7 +32,7 @@ repos:
           additional_dependencies: ["bandit[toml]"]
 
   -   repo: https://github.com/astral-sh/ruff-pre-commit
-      rev: v0.8.2
+      rev: v0.8.6
       hooks:
       -   id: ruff
           args: [--fix, --exit-non-zero-on-fix]

CONTRIBUTING.md

@@ -155,31 +155,20 @@ Before starting your work on the project, set up your development environment:
     source .venv/bin/activate
     ```
 
-3. Install Poetry:
+3. Install `uv`:
 
-    Using pip:
-
-    ```bash
-    pip install -U pip setuptools
-    pip install poetry
-    ```
-
-    Or using pipx (recommended for global installation):
-
-    ```bash
-    pipx install poetry
-    ```
+    Follow the instructions on the [uv installation page](https://docs.astral.sh/uv/getting-started/installation/).
 
 4. Install project dependencies:
 
     ```bash
-    poetry install
+    uv pip install -r pyproject.toml --extra dev --extra docs --extra metrics
     ```
 
 5. Run pytest to verify the setup:
 
     ```bash
-    poetry run pytest
+    uv run pytest
     ```
 
 ## 🎨 Code Style and Quality
@@ -192,7 +181,7 @@ 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 --with dev`. 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: `uv pip install -r pyproject.toml --extra 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.
 
@@ -214,7 +203,7 @@ So far, **there is no type checking with mypy**. See [issue](https://github.com/
 
 The `supervision` documentation is stored in a folder called `docs`. The project documentation is built using `mkdocs`.
 
-To run the documentation, install the project requirements with `poetry install --with dev`. Then, run `mkdocs serve` to start the documentation server.
+To run the documentation, install the project requirements with `uv pip install -r pyproject.toml --extra dev --extra docs`. Then, run `mkdocs serve` to start the documentation server.
 
 You can learn more about mkdocs on the [mkdocs website](https://www.mkdocs.org/).
 

docs/detection/utils.md

@@ -89,6 +89,12 @@ status: new
 
 :::supervision.detection.utils.xywh_to_xyxy
 
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.xyxy_to_xywh">xyxy_to_xywh</a></h2>
+</div>
+
+:::supervision.detection.utils.xyxy_to_xywh
+
 <div class="md-typeset">
   <h2><a href="#supervision.detection.utils.xcycwh_to_xyxy">xcycwh_to_xyxy</a></h2>
 </div>

docs/how_to/track_objects.md

@@ -55,7 +55,7 @@ it will be modified to include tracking, labeling, and trace annotations.
     from ultralytics import YOLO
 
     model = YOLO("yolov8n.pt")
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
 
     def callback(frame: np.ndarray, _: int) -> np.ndarray:
         results = model(frame)[0]
@@ -77,7 +77,7 @@ it will be modified to include tracking, labeling, and trace annotations.
     from inference.models.utils import get_roboflow_model
 
     model = get_roboflow_model(model_id="yolov8n-640", api_key=<ROBOFLOW API KEY>)
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
 
     def callback(frame: np.ndarray, _: int) -> np.ndarray:
         results = model.infer(frame)[0]
@@ -112,7 +112,7 @@ enabling the continuous following of the object's motion path across different f
 
     model = YOLO("yolov8n.pt")
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
 
     def callback(frame: np.ndarray, _: int) -> np.ndarray:
         results = model(frame)[0]
@@ -136,7 +136,7 @@ enabling the continuous following of the object's motion path across different f
 
     model = get_roboflow_model(model_id="yolov8n-640", api_key=<ROBOFLOW API KEY>)
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
 
     def callback(frame: np.ndarray, _: int) -> np.ndarray:
         results = model.infer(frame)[0]
@@ -168,7 +168,7 @@ offering a clear visual representation of each object's class and unique identif
 
     model = YOLO("yolov8n.pt")
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     def callback(frame: np.ndarray, _: int) -> np.ndarray:
@@ -203,7 +203,7 @@ offering a clear visual representation of each object's class and unique identif
 
     model = get_roboflow_model(model_id="yolov8n-640", api_key=<ROBOFLOW API KEY>)
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
 
     def callback(frame: np.ndarray, _: int) -> np.ndarray:
@@ -250,7 +250,7 @@ movement patterns and interactions between objects in the video.
 
     model = YOLO("yolov8n.pt")
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
     trace_annotator = sv.TraceAnnotator()
 
@@ -288,7 +288,7 @@ movement patterns and interactions between objects in the video.
 
     model = get_roboflow_model(model_id="yolov8n-640", api_key=<ROBOFLOW API KEY>)
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
     trace_annotator = sv.TraceAnnotator()
 

docs/index.md

@@ -113,18 +113,19 @@ You can install `supervision` in a
         pip install -e "."
         ```
 
-    === "poetry"
+    === "uv"
         ```bash
         # clone repository and navigate to root directory
         git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
         cd supervision
 
         # setup python environment and activate it
-        poetry env use python3.10
-        poetry shell
+        uv venv
+        source .venv/bin/activate
 
         # installation
-        poetry install
+        uv pip install -r pyproject.toml -e . --all-extras
+
         ```
 
 ## 🚀 Quickstart

examples/count_people_in_zone/inference_example.py

@@ -35,9 +35,7 @@ def load_zones_config(file_path: str) -> List[np.ndarray]:
 
 def initiate_annotators(
     polygons: List[np.ndarray], resolution_wh: Tuple[int, int]
-) -> Tuple[
-    List[sv.PolygonZone], List[sv.PolygonZoneAnnotator], List[sv.BoundingBoxAnnotator]
-]:
+) -> Tuple[List[sv.PolygonZone], List[sv.PolygonZoneAnnotator], List[sv.BoxAnnotator]]:
     line_thickness = sv.calculate_optimal_line_thickness(resolution_wh=resolution_wh)
     text_scale = sv.calculate_optimal_text_scale(resolution_wh=resolution_wh)
 
@@ -54,7 +52,7 @@ def initiate_annotators(
             text_thickness=line_thickness * 2,
             text_scale=text_scale * 2,
         )
-        box_annotator = sv.BoundingBoxAnnotator(
+        box_annotator = sv.BoxAnnotator(
             color=COLORS.by_idx(index), thickness=line_thickness
         )
         zones.append(zone)
@@ -97,7 +95,7 @@ def annotate(
     frame: np.ndarray,
     zones: List[sv.PolygonZone],
     zone_annotators: List[sv.PolygonZoneAnnotator],
-    box_annotators: List[sv.BoundingBoxAnnotator],
+    box_annotators: List[sv.BoxAnnotator],
     detections: sv.Detections,
 ) -> np.ndarray:
     """
@@ -108,7 +106,7 @@ def annotate(
         zones (List[sv.PolygonZone]): A list of polygon zones used for detection.
         zone_annotators (List[sv.PolygonZoneAnnotator]): A list of annotators for
             drawing zone annotations.
-        box_annotators (List[sv.BoundingBoxAnnotator]): A list of annotators for
+        box_annotators (List[sv.BoxAnnotator]): A list of annotators for
             drawing box annotations.
         detections (sv.Detections): Detections to be used for annotation.
 

examples/count_people_in_zone/ultralytics_example.py

@@ -33,9 +33,7 @@ def load_zones_config(file_path: str) -> List[np.ndarray]:
 
 def initiate_annotators(
     polygons: List[np.ndarray], resolution_wh: Tuple[int, int]
-) -> Tuple[
-    List[sv.PolygonZone], List[sv.PolygonZoneAnnotator], List[sv.BoundingBoxAnnotator]
-]:
+) -> Tuple[List[sv.PolygonZone], List[sv.PolygonZoneAnnotator], List[sv.BoxAnnotator]]:
     line_thickness = sv.calculate_optimal_line_thickness(resolution_wh=resolution_wh)
     text_scale = sv.calculate_optimal_text_scale(resolution_wh=resolution_wh)
 
@@ -52,7 +50,7 @@ def initiate_annotators(
             text_thickness=line_thickness * 2,
             text_scale=text_scale * 2,
         )
-        box_annotator = sv.BoundingBoxAnnotator(
+        box_annotator = sv.BoxAnnotator(
             color=COLORS.by_idx(index), thickness=line_thickness
         )
         zones.append(zone)
@@ -94,7 +92,7 @@ def annotate(
     frame: np.ndarray,
     zones: List[sv.PolygonZone],
     zone_annotators: List[sv.PolygonZoneAnnotator],
-    box_annotators: List[sv.BoundingBoxAnnotator],
+    box_annotators: List[sv.BoxAnnotator],
     detections: sv.Detections,
 ) -> np.ndarray:
     """
@@ -105,7 +103,7 @@ def annotate(
         zones (List[sv.PolygonZone]): A list of polygon zones used for detection.
         zone_annotators (List[sv.PolygonZoneAnnotator]): A list of annotators for
             drawing zone annotations.
-        box_annotators (List[sv.BoundingBoxAnnotator]): A list of annotators for
+        box_annotators (List[sv.BoxAnnotator]): A list of annotators for
             drawing box annotations.
         detections (sv.Detections): Detections to be used for annotation.
 

examples/tracking/inference_example.py

@@ -18,7 +18,7 @@ def process_video(
     model = get_roboflow_model(model_id=model_id, api_key=roboflow_api_key)
 
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
     frame_generator = sv.get_video_frames_generator(source_path=source_video_path)
     video_info = sv.VideoInfo.from_video_path(video_path=source_video_path)

examples/tracking/ultralytics_example.py

@@ -16,7 +16,7 @@ def process_video(
     model = YOLO(source_weights_path)
 
     tracker = sv.ByteTrack()
-    box_annotator = sv.BoundingBoxAnnotator()
+    box_annotator = sv.BoxAnnotator()
     label_annotator = sv.LabelAnnotator()
     frame_generator = sv.get_video_frames_generator(source_path=source_video_path)
     video_info = sv.VideoInfo.from_video_path(video_path=source_video_path)

mkdocs.yml

@@ -178,8 +178,7 @@ markdown_extensions:
       anchor_linenums: true
 
 extra_javascript:
-  # disabled until 0.27.0, so we can test its functionality
-  # - "javascripts/init_kapa_widget.js"
+  - "javascripts/init_kapa_widget.js"
   - "javascripts/cookbooks-card.js"
   - "https://cdnjs.cloudflare.com/ajax/libs/dompurify/3.0.8/purify.min.js"
 

pyproject.toml

@@ -1,18 +1,16 @@
-[tool.poetry]
+[project]
 name = "supervision"
-version = "0.26.0rc2"
 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>",
-    "Linas Kondrackis <linas@roboflow.com>",
-]
+license = { text = "MIT" }
+version = "0.26.0rc4"
 readme = "README.md"
-license = "MIT"
-packages = [{ include = "supervision" }, { include = "supervision/py.typed" }]
-homepage = "https://github.com/roboflow/supervision"
-repository = "https://github.com/roboflow/supervision"
-documentation = "https://supervision.roboflow.com/latest/"
+requires-python = ">=3.8"
+authors = [
+    { name = "Piotr Skalski", email = "piotr.skalski92@gmail.com" }
+]
+maintainers = [
+    { name = "Piotr Skalski", email = "piotr.skalski92@gmail.com" },
+]
 keywords = [
     "machine-learning",
     "deep-learning",
@@ -22,7 +20,6 @@ keywords = [
     "AI",
     "Roboflow",
 ]
-
 classifiers = [
     'Development Status :: 4 - Beta',
     'Intended Audience :: Developers',
@@ -40,116 +37,59 @@ classifiers = [
     'Operating System :: POSIX :: Linux',
     'Operating System :: MacOS',
 ]
-
-[tool.poetry.dependencies]
-python = "^3.8"
-
-# For Python versions 3.13 and above, use numpy versions 2.1.0,
-# required for poetry to not fail.
-numpy = [
-    { version = ">=1.21.2", python = "<3.13" },
-    { version = ">=2.1.0", python = ">=3.13" },
+dependencies = [
+    "numpy>=1.21.2",
+    "scipy>=1.10.0",
+    "matplotlib>=3.6.0",
+    "pyyaml>=5.3",
+    "defusedxml>=0.7.1",
+    "pillow>=9.4",
+    "requests>=2.26.0",
+    "tqdm>=4.62.3",
+    "opencv-python>=4.5.5.64"
 ]
 
-scipy = [
-    { version = "1.10.0", python = "<3.9" },
-    { version = "^1.10.0", python = ">=3.9" },
-    { version = ">=1.14.1", python = ">=3.13" },
+[project.urls]
+Homepage = "https://github.com/roboflow/supervision"
+Repository = "https://github.com/roboflow/supervision"
+Documentation = "https://supervision.roboflow.com/latest/"
 
+[project.optional-dependencies]
+metrics = [
+    "pandas>=2.0.0",
 ]
-
-# Matplotlib sub-dependency
-# The 'contourpy' package is required by Matplotlib for contour plotting.
-# We need to ensure compatibility with both Python 3.8 and Python 3.13.
-#
-# For Python 3.8 and above, we use version 1.0.7 or higher, as it is the lowest major version that supports Python 3.8.
-# For Python 3.13 and above, we use version 1.3.0 or higher, as it is the first version that explicitly supports Python 3.13.
-contourpy = [
-    { version = ">=1.0.7", python = ">=3.8,<3.13" },
-    { version = ">=1.3.0", python = ">=3.13" },
+dev = [
+    "pytest>=7.2.2,<9.0.0",
+    "tox>=4.11.4",
+    "notebook>=6.5.3,<8.0.0",
+    "ipywidgets>=8.1.1",
+    "jupytext>=1.16.1",
+    "nbconvert>=7.14.2",
+    "docutils!=0.21"
 ]
-
-matplotlib = [
-    { version = ">=3.6.0,<3.8.0", python = "3.8" },
-    { version = ">=3.6.0", python = ">=3.9" },
-    { version = ">=3.7.3", python = ">=3.12" },
-    { version = ">=3.9.2", python = ">=3.13" },
+docs = [
+    "mkdocs-material[imaging]>=9.5.5",
+    "mkdocstrings>=0.25.2,<0.27.0",
+    "mkdocstrings-python>=1.10.9",
+    "mike>=2.0.0",
+    "mkdocs-jupyter>=0.24.3",
+    "mkdocs-git-committers-plugin-2>=2.4.1; python_version >= '3.8' and python_version < '4'",
+    "mkdocs-git-revision-date-localized-plugin>=1.2.4"
 ]
-
-
-pyyaml = ">=5.3"
-defusedxml = "^0.7.1"
-pillow = ">=9.4"
-requests = ">=2.26.0"
-tqdm = ">=4.62.3"
-# pandas: picked lowest major version that supports Python 3.8
-# pandas 2.2.3 has been released with support for Python 3.13
-pandas = [
-    { version = ">=2.0.0", python = "<3.13", optional = true },
-    { version = ">=2.2.3", python = ">=3.13", optional = true },
+build = [
+    "twine>=5.1.1,<7.0.0",
+    "wheel>=0.40,<0.46",
+    "build>=0.10,<1.3"
 ]
 
-opencv-python = ">=4.5.5.64"
-
-[tool.poetry.extras]
-metrics = ["pandas"]
-
-[tool.poetry.group.test.dependencies]
-pytest = ">=7.2.2,<9.0.0"
-pytest-md = "^0.2.0"
-pytest-emoji = "^0.2.0"
-
-[tool.poetry.group.dev.dependencies]
-twine = ">=5.1.1,<7.0.0"
-pytest = ">=7.2.2,<9.0.0"
-wheel = ">=0.40,<0.46"
-build = ">=0.10,<1.3"
-ruff = ">=0.1.0"
-mypy = "^1.4.1"
-pre-commit = "^3.3.3"
-tox = "^4.11.4"
-notebook = ">=6.5.3,<8.0.0"
-ipywidgets = "^8.1.1"
-jupytext = "^1.16.1"
-nbconvert = "^7.14.2"
-docutils = [
-    { version = "^0.20.1", python = "<3.9" },
-    { version = "^0.21.1", python = ">=3.9" },
-]
-
-[tool.poetry.group.docs.dependencies]
-mkdocs-material = { extras = ["imaging"], version = "^9.5.5" }
-mkdocstrings = ">=0.25.2,<0.27.0"
-mkdocstrings-python = "^1.10.9"
-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
-mkdocs-jupyter = "^0.24.3"
-mkdocs-git-committers-plugin-2 = "^2.4.1"
-mkdocs-git-revision-date-localized-plugin = "^1.2.4"
-
-[tool.poetry.group.typecheck]
-optional = true
-
-[tool.poetry.group.typecheck.dependencies]
-types-pyyaml = "^6.0.12.20240808"
-types-cffi = "^1.16.0.20240331"
-types-requests = "^2.32.0.20240712"
-types-tqdm = "^4.66.0.20240417"
-pandas-stubs = ">=2.0.0.230412"
-
-[tool.poetry.group.build.dependencies]
-twine = ">=5.1.1,<7.0.0"
-
 [tool.bandit]
 target = ["test", "supervision"]
-tests = ["B201", "B301", "B318", "B314", "B303", "B413", "B412", "B410"]
+tests = ["B201", "B301", "B318", "B314", "B303", "B413", "B412"]
 
 [tool.autoflake]
 check = true
 imports = ["cv2", "supervision"]
 
-
 [tool.ruff]
 target-version = "py38"
 
@@ -273,7 +213,7 @@ skip-magic-trailing-comma = false
 line-ending = "auto"
 
 [tool.codespell]
-skip = "*.ipynb,poetry.lock"
+skip = "*.ipynb"
 count = true
 quiet-level = 3
 ignore-words-list = "STrack,sTrack,strack"
@@ -282,8 +222,12 @@ ignore-words-list = "STrack,sTrack,strack"
 include-package-data = false
 
 [tool.setuptools.packages.find]
+include = ["supervision*"]
 exclude = ["docs*", "test*", "examples*"]
 
+[tool.setuptools.package-data]
+supervision = ["py.typed"]
+
 [build-system]
-requires = ["poetry-core"]
-build-backend = "poetry.core.masonry.api"
+requires = ["setuptools >= 61.0"]
+build-backend = "setuptools.build_meta"

supervision/__init__.py

@@ -9,7 +9,6 @@ except importlib_metadata.PackageNotFoundError:
 from supervision.annotators.core import (
     BackgroundOverlayAnnotator,
     BlurAnnotator,
-    BoundingBoxAnnotator,
     BoxAnnotator,
     BoxCornerAnnotator,
     CircleAnnotator,
@@ -46,7 +45,6 @@ from supervision.detection.line_zone import (
     LineZoneAnnotator,
     LineZoneAnnotatorMulticlass,
 )
-from supervision.detection.lmm import LMM
 from supervision.detection.overlap_filter import (
     OverlapFilter,
     box_non_max_merge,
@@ -78,7 +76,9 @@ from supervision.detection.utils import (
     xcycwh_to_xyxy,
     xywh_to_xyxy,
     xyxy_to_polygons,
+    xyxy_to_xywh,
 )
+from supervision.detection.vlm import LMM, VLM
 from supervision.draw.color import Color, ColorPalette
 from supervision.draw.utils import (
     calculate_optimal_line_thickness,
@@ -126,7 +126,6 @@ __all__ = [
     "BackgroundOverlayAnnotator",
     "BaseDataset",
     "BlurAnnotator",
-    "BoundingBoxAnnotator",
     "BoxAnnotator",
     "BoxCornerAnnotator",
     "ByteTrack",
@@ -226,4 +225,5 @@ __all__ = [
     "xcycwh_to_xyxy",
     "xywh_to_xyxy",
     "xyxy_to_polygons",
+    "xyxy_to_xywh",
 ]

supervision/annotators/core.py

@@ -36,7 +36,6 @@ from supervision.utils.image import (
     overlay_image,
     scale_image,
 )
-from supervision.utils.internal import deprecated
 
 CV2_FONT = cv2.FONT_HERSHEY_SIMPLEX
 
@@ -124,92 +123,6 @@ class BoxAnnotator(BaseAnnotator):
         return scene
 
 
-@deprecated(
-    "`BoundingBoxAnnotator` is deprecated and has been renamed to `BoxAnnotator`."
-    " `BoundingBoxAnnotator` will be removed in supervision-0.26.0."
-)
-class BoundingBoxAnnotator(BaseAnnotator):
-    """
-    A class for drawing bounding boxes on an image using provided detections.
-    """
-
-    def __init__(
-        self,
-        color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
-        thickness: int = 2,
-        color_lookup: ColorLookup = ColorLookup.CLASS,
-    ):
-        """
-        Args:
-            color (Union[Color, ColorPalette]): The color or color palette to use for
-                annotating detections.
-            thickness (int): Thickness of the bounding box lines.
-            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
-                Options are `INDEX`, `CLASS`, `TRACK`.
-        """
-        self.color: Union[Color, ColorPalette] = color
-        self.thickness: int = thickness
-        self.color_lookup: ColorLookup = color_lookup
-
-    @ensure_cv2_image_for_annotation
-    def annotate(
-        self,
-        scene: ImageType,
-        detections: Detections,
-        custom_color_lookup: Optional[np.ndarray] = None,
-    ) -> ImageType:
-        """
-        Annotates the given scene with bounding boxes based on the provided detections.
-
-        Args:
-            scene (ImageType): The image where bounding boxes will be drawn. `ImageType`
-            is a flexible type, accepting either `numpy.ndarray` or `PIL.Image.Image`.
-            detections (Detections): Object detections to annotate.
-            custom_color_lookup (Optional[np.ndarray]): Custom color lookup array.
-                Allows to override the default color mapping strategy.
-
-        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(...)
-
-            bounding_box_annotator = sv.BoundingBoxAnnotator()
-            annotated_frame = bounding_box_annotator.annotate(
-                scene=image.copy(),
-                detections=detections
-            )
-            ```
-
-        ![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(
-                color=self.color,
-                detections=detections,
-                detection_idx=detection_idx,
-                color_lookup=self.color_lookup
-                if custom_color_lookup is None
-                else custom_color_lookup,
-            )
-            cv2.rectangle(
-                img=scene,
-                pt1=(x1, y1),
-                pt2=(x2, y2),
-                color=color.as_bgr(),
-                thickness=self.thickness,
-            )
-        return scene
-
-
 class OrientedBoxAnnotator(BaseAnnotator):
     """
     A class for drawing oriented bounding boxes on an image using provided detections.

supervision/dataset/core.py

@@ -705,28 +705,6 @@ class ClassificationDataset(BaseDataset):
                 "a list of paths `List[str]` instead."
             )
 
-    @property
-    @deprecated(
-        "`DetectionDataset.images` property is deprecated and will be removed in "
-        "`supervision-0.26.0`. Iterate with `for path, image, annotation in dataset:` "
-        "instead."
-    )
-    def images(self) -> Dict[str, np.ndarray]:
-        """
-        Load all images to memory and return them as a dictionary.
-
-        !!! warning
-
-            Only use this when you need all images at once.
-            It is much more memory-efficient to initialize dataset with
-            image paths and use `for path, image, annotation in dataset:`.
-        """
-        if self._images_in_memory:
-            return self._images_in_memory
-
-        images = {image_path: cv2.imread(image_path) for image_path in self.image_paths}
-        return images
-
     def _get_image(self, image_path: str) -> np.ndarray:
         """Assumes that image is in dataset"""
         if self._images_in_memory:

supervision/dataset/formats/yolo.py

@@ -2,8 +2,8 @@ import os
 from pathlib import Path
 from typing import TYPE_CHECKING, Dict, List, Optional, Tuple
 
-import cv2
 import numpy as np
+from PIL import Image
 
 from supervision.config import ORIENTED_BOX_COORDINATES
 from supervision.dataset.utils import approximate_mask_with_polygons
@@ -153,7 +153,18 @@ def load_yolo_annotations(
     image_paths = [
         str(path)
         for path in list_files_with_extensions(
-            directory=images_directory_path, extensions=["jpg", "jpeg", "png"]
+            directory=images_directory_path,
+            extensions=[
+                "bmp",
+                "dng",
+                "jpg",
+                "jpeg",
+                "mpo",
+                "png",
+                "tif",
+                "tiff",
+                "webp",
+            ],
         )
     ]
 
@@ -167,10 +178,16 @@ def load_yolo_annotations(
             annotations[image_path] = Detections.empty()
             continue
 
-        image = cv2.imread(image_path)
+        # PIL is much faster than cv2 for checking image shape and mode: https://github.com/roboflow/supervision/issues/1554
+        image = Image.open(image_path)
         lines = read_txt_file(file_path=annotation_path, skip_empty=True)
-        h, w, _ = image.shape
+        w, h = image.size
         resolution_wh = (w, h)
+        if image.mode not in ("RGB", "L"):
+            raise ValueError(
+                f"Images must be 'RGB' or 'grayscale', \
+                but {image_path} mode is '{image.mode}'."
+            )
 
         with_masks = _with_mask(lines=lines)
         with_masks = force_masks if force_masks else with_masks

supervision/detection/core.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 from dataclasses import dataclass, field
+from enum import Enum
 from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
 
 import numpy as np
@@ -9,12 +10,6 @@ from supervision.config import (
     CLASS_NAME_DATA_FIELD,
     ORIENTED_BOX_COORDINATES,
 )
-from supervision.detection.lmm import (
-    LMM,
-    from_florence_2,
-    from_paligemma,
-    validate_lmm_parameters,
-)
 from supervision.detection.overlap_filter import (
     box_non_max_merge,
     box_non_max_suppression,
@@ -38,8 +33,16 @@ from supervision.detection.utils import (
     process_roboflow_result,
     xywh_to_xyxy,
 )
+from supervision.detection.vlm import (
+    LMM,
+    VLM,
+    from_florence_2,
+    from_paligemma,
+    from_qwen_2_5_vl,
+    validate_vlm_parameters,
+)
 from supervision.geometry.core import Position
-from supervision.utils.internal import get_instance_variables
+from supervision.utils.internal import deprecated, get_instance_variables
 from supervision.validators import validate_detections_fields
 
 
@@ -799,6 +802,10 @@ class Detections:
         )
 
     @classmethod
+    @deprecated(
+        "`Detections.from_lmm` property is deprecated and will be removed in "
+        "`supervision-0.31.0`. Use Detections.from_vlm instead."
+    )
     def from_lmm(
         cls, lmm: Union[LMM, str], result: Union[str, dict], **kwargs: Any
     ) -> Detections:
@@ -837,16 +844,51 @@ class Detections:
             # array([0])
             ```
         """
-        lmm = validate_lmm_parameters(lmm, result, kwargs)
+        # filler logic mapping old from_lmm to new from_vlm
+        lmm_to_vlm = {
+            LMM.PALIGEMMA: VLM.PALIGEMMA,
+            LMM.FLORENCE_2: VLM.FLORENCE_2,
+            LMM.QWEN_2_5_VL: VLM.QWEN_2_5_VL,
+        }
 
-        if lmm == LMM.PALIGEMMA:
-            assert isinstance(result, str)
+        # (this works even if the LMM enum is wrapped by @deprecated)
+        if isinstance(lmm, Enum) and lmm.__class__.__name__ == "LMM":
+            vlm = lmm_to_vlm[lmm]
+
+        elif isinstance(lmm, str):
+            try:
+                lmm_enum = LMM(lmm.lower())
+            except ValueError:
+                raise ValueError(
+                    f"Invalid LMM string '{lmm}'. Must be one of "
+                    f"{[m.value for m in LMM]}"
+                )
+            vlm = lmm_to_vlm[lmm_enum]
+
+        else:
+            raise ValueError(
+                f"Invalid type for 'lmm': {type(lmm)}. Must be LMM or str."
+            )
+
+        return cls.from_vlm(vlm=vlm, result=result, **kwargs)
+
+    @classmethod
+    def from_vlm(
+        cls, vlm: Union[VLM, str], result: Union[str, dict], **kwargs: Any
+    ) -> Detections:
+        vlm = validate_vlm_parameters(vlm, result, kwargs)
+
+        if vlm == VLM.PALIGEMMA:
             xyxy, class_id, class_name = from_paligemma(result, **kwargs)
             data = {CLASS_NAME_DATA_FIELD: class_name}
             return cls(xyxy=xyxy, class_id=class_id, data=data)
 
-        if lmm == LMM.FLORENCE_2:
-            assert isinstance(result, dict)
+        if vlm == VLM.QWEN_2_5_VL:
+            xyxy, class_id, class_name = from_qwen_2_5_vl(result, **kwargs)
+            data = {CLASS_NAME_DATA_FIELD: class_name}
+            return cls(xyxy=xyxy, class_id=class_id, data=data)
+
+        if vlm == VLM.FLORENCE_2:
             xyxy, labels, mask, xyxyxyxy = from_florence_2(result, **kwargs)
             if len(xyxy) == 0:
                 return cls.empty()
@@ -859,8 +901,6 @@ class Detections:
 
             return cls(xyxy=xyxy, mask=mask, data=data)
 
-        raise ValueError(f"Unsupported LMM: {lmm}")
-
     @classmethod
     def from_easyocr(cls, easyocr_results: list) -> Detections:
         """
@@ -1027,16 +1067,16 @@ class Detections:
             detections_1 = sv.Detections(
                 xyxy=np.array([[15, 15, 100, 100], [200, 200, 300, 300]]),
                 class_id=np.array([1, 2]),
-                data={'feature_vector': np.array([0.1, 0.2)])}
-             )
+                data={'feature_vector': np.array([0.1, 0.2])}
+            )
 
             detections_2 = sv.Detections(
                 xyxy=np.array([[30, 30, 120, 120]]),
                 class_id=np.array([1]),
-                data={'feature_vector': [np.array([0.3])]}
-             )
+                data={'feature_vector': np.array([0.3])}
+            )
 
-            merged_detections = Detections.merge([detections_1, detections_2])
+            merged_detections = sv.Detections.merge([detections_1, detections_2])
 
             merged_detections.xyxy
             array([[ 15,  15, 100, 100],
@@ -1203,7 +1243,7 @@ class Detections:
         if isinstance(index, str):
             return self.data.get(index)
         if self.is_empty():
-            return Detections.empty()
+            return self
         if isinstance(index, int):
             index = [index]
         return Detections(

supervision/detection/utils.py

@@ -321,6 +321,43 @@ def xywh_to_xyxy(xywh: np.ndarray) -> np.ndarray:
     return xyxy
 
 
+def xyxy_to_xywh(xyxy: np.ndarray) -> np.ndarray:
+    """
+    Converts bounding box coordinates from `(x_min, y_min, x_max, y_max)`
+    format to `(x, y, width, height)` format.
+
+    Args:
+        xyxy (np.ndarray): A numpy array of shape `(N, 4)` where each row
+            corresponds to a bounding box in the format `(x_min, y_min, x_max,
+            y_max)`.
+
+    Returns:
+        np.ndarray: A numpy array of shape `(N, 4)` where each row corresponds
+            to a bounding box in the format `(x, y, width, height)`.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xyxy = np.array([
+            [10, 20, 40, 60],
+            [15, 25, 50, 70]
+        ])
+
+        sv.xyxy_to_xywh(xyxy=xyxy)
+        # array([
+        #     [10, 20, 30, 40],
+        #     [15, 25, 35, 45]
+        # ])
+        ```
+    """
+    xywh = xyxy.copy()
+    xywh[:, 2] = xyxy[:, 2] - xyxy[:, 0]
+    xywh[:, 3] = xyxy[:, 3] - xyxy[:, 1]
+    return xywh
+
+
 def xcycwh_to_xyxy(xcycwh: np.ndarray) -> np.ndarray:
     """
     Converts bounding box coordinates from `(center_x, center_y, width, height)`

supervision/detection/vlm.py

@@ -1,3 +1,4 @@
+import json
 import re
 from enum import Enum
 from typing import Any, Dict, List, Optional, Tuple, Union
@@ -5,23 +6,41 @@ from typing import Any, Dict, List, Optional, Tuple, Union
 import numpy as np
 
 from supervision.detection.utils import polygon_to_mask, polygon_to_xyxy
+from supervision.utils.internal import deprecated
 
 
+@deprecated(
+    "`LMM` enum is deprecated and will be removed in "
+    "`supervision-0.31.0`. Use VLM instead."
+)
 class LMM(Enum):
     PALIGEMMA = "paligemma"
     FLORENCE_2 = "florence_2"
+    QWEN_2_5_VL = "qwen_2_5_vl"
 
 
-RESULT_TYPES: Dict[LMM, type] = {LMM.PALIGEMMA: str, LMM.FLORENCE_2: dict}
+class VLM(Enum):
+    PALIGEMMA = "paligemma"
+    FLORENCE_2 = "florence_2"
+    QWEN_2_5_VL = "qwen_2_5_vl"
 
-REQUIRED_ARGUMENTS: Dict[LMM, List[str]] = {
-    LMM.PALIGEMMA: ["resolution_wh"],
-    LMM.FLORENCE_2: ["resolution_wh"],
+
+RESULT_TYPES: Dict[VLM, type] = {
+    VLM.PALIGEMMA: str,
+    VLM.FLORENCE_2: dict,
+    VLM.QWEN_2_5_VL: str,
 }
 
-ALLOWED_ARGUMENTS: Dict[LMM, List[str]] = {
-    LMM.PALIGEMMA: ["resolution_wh", "classes"],
-    LMM.FLORENCE_2: ["resolution_wh"],
+REQUIRED_ARGUMENTS: Dict[VLM, List[str]] = {
+    VLM.PALIGEMMA: ["resolution_wh"],
+    VLM.FLORENCE_2: ["resolution_wh"],
+    VLM.QWEN_2_5_VL: ["input_wh", "resolution_wh"],
+}
+
+ALLOWED_ARGUMENTS: Dict[VLM, List[str]] = {
+    VLM.PALIGEMMA: ["resolution_wh", "classes"],
+    VLM.FLORENCE_2: ["resolution_wh"],
+    VLM.QWEN_2_5_VL: ["input_wh", "resolution_wh", "classes"],
 }
 
 SUPPORTED_TASKS_FLORENCE_2 = [
@@ -38,58 +57,172 @@ SUPPORTED_TASKS_FLORENCE_2 = [
 ]
 
 
-def validate_lmm_parameters(
-    lmm: Union[LMM, str], result: Any, kwargs: Dict[str, Any]
-) -> LMM:
-    if isinstance(lmm, str):
+def validate_vlm_parameters(
+    vlm: Union[VLM, str], result: Any, kwargs: Dict[str, Any]
+) -> VLM:
+    if isinstance(vlm, str):
         try:
-            lmm = LMM(lmm.lower())
+            vlm = VLM(vlm.lower())
         except ValueError:
             raise ValueError(
-                f"Invalid lmm value: {lmm}. Must be one of {[e.value for e in LMM]}"
+                f"Invalid vlm value: {vlm}. Must be one of {[e.value for e in VLM]}"
             )
 
-    if not isinstance(result, RESULT_TYPES[lmm]):
+    if not isinstance(result, RESULT_TYPES[vlm]):
         raise ValueError(
-            f"Invalid LMM result type: {type(result)}. Must be {RESULT_TYPES[lmm]}"
+            f"Invalid VLM result type: {type(result)}. Must be {RESULT_TYPES[vlm]}"
         )
 
-    required_args = REQUIRED_ARGUMENTS.get(lmm, [])
+    required_args = REQUIRED_ARGUMENTS.get(vlm, [])
     for arg in required_args:
         if arg not in kwargs:
             raise ValueError(f"Missing required argument: {arg}")
 
-    allowed_args = ALLOWED_ARGUMENTS.get(lmm, [])
+    allowed_args = ALLOWED_ARGUMENTS.get(vlm, [])
     for arg in kwargs:
         if arg not in allowed_args:
-            raise ValueError(f"Argument {arg} is not allowed for {lmm.name}")
+            raise ValueError(f"Argument {arg} is not allowed for {vlm.name}")
 
-    return lmm
+    return vlm
 
 
 def from_paligemma(
     result: str, resolution_wh: Tuple[int, int], classes: Optional[List[str]] = None
 ) -> Tuple[np.ndarray, Optional[np.ndarray], np.ndarray]:
+    """
+    Parse bounding boxes from paligemma-formatted text, scale them to the specified
+    resolution, and optionally filter by classes.
+
+    Args:
+        result: String containing paligemma-formatted locations and labels.
+        resolution_wh: Tuple (width, height) to which we scale the box coordinates.
+        classes: Optional list of valid class names. If provided, boxes and labels not
+            in this list are filtered out.
+
+    Returns:
+        xyxy (np.ndarray): An array of shape `(n, 4)` containing
+            the bounding boxes coordinates in format `[x1, y1, x2, y2]`.
+        class_id (Optional[np.ndarray]): An array of shape `(n,)` containing
+            the class indices for each bounding box (or `None` if classes is not
+            provided).
+        class_name (np.ndarray): An array of shape `(n,)` containing
+            the class labels for each bounding box.
+    """
+
     w, h = resolution_wh
+    if w <= 0 or h <= 0:
+        raise ValueError(
+            f"Both dimensions in resolution_wh must be positive. Got ({w}, {h})."
+        )
+
     pattern = re.compile(
         r"(?<!<loc\d{4}>)<loc(\d{4})><loc(\d{4})><loc(\d{4})><loc(\d{4})> ([\w\s\-]+)"
     )
     matches = pattern.findall(result)
     matches = np.array(matches) if matches else np.empty((0, 5))
 
+    if matches.shape[0] == 0:
+        return np.empty((0, 4)), None, np.empty(0, dtype=str)
+
     xyxy, class_name = matches[:, [1, 0, 3, 2]], matches[:, 4]
     xyxy = xyxy.astype(int) / 1024 * np.array([w, h, w, h])
     class_name = np.char.strip(class_name.astype(str))
     class_id = None
 
     if classes is not None:
-        mask = np.array([name in classes for name in class_name]).astype(bool)
-        xyxy, class_name = xyxy[mask], class_name[mask]
+        mask = np.array([name in classes for name in class_name], dtype=bool)
+        xyxy = xyxy[mask]
+        class_name = class_name[mask]
         class_id = np.array([classes.index(name) for name in class_name])
 
     return xyxy, class_id, class_name
 
 
+def from_qwen_2_5_vl(
+    result: str,
+    input_wh: Tuple[int, int],
+    resolution_wh: Tuple[int, int],
+    classes: Optional[List[str]] = None,
+) -> Tuple[np.ndarray, Optional[np.ndarray], np.ndarray]:
+    """
+    Parse and scale bounding boxes from Qwen-2.5-VL style JSON output.
+
+    The JSON is expected to be enclosed in triple backticks with the format:
+      ```json
+      [
+          {"bbox_2d": [x1, y1, x2, y2], "label": "some class name"},
+          ...
+      ]
+      ```
+
+    Args:
+        result: String containing the JSON snippet enclosed by triple backticks.
+        input_wh: (input_width, input_height) describing the original bounding box
+            scale.
+        resolution_wh: (output_width, output_height) to which we rescale the boxes.
+        classes: Optional list of valid class names. If provided, returned boxes/labels
+            are filtered to only those classes found here.
+
+    Returns:
+        xyxy (np.ndarray): An array of shape `(n, 4)` containing
+            the bounding boxes coordinates in format `[x1, y1, x2, y2]`
+        class_id (Optional[np.ndarray]): An array of shape `(n,)` containing
+            the class indices for each bounding box (or None if `classes` is not
+            provided)
+        class_name (np.ndarray): An array of shape `(n,)` containing
+            the class labels for each bounding box
+    """
+    in_w, in_h = input_wh
+    out_w, out_h = resolution_wh
+
+    if in_w <= 0 or in_h <= 0 or out_w <= 0 or out_h <= 0:
+        raise ValueError(
+            f"Both input and resolution dimensions must be positive. "
+            f"Got input_wh=({in_w}, {in_h}), resolution_wh=({out_w}, {out_h})."
+        )
+
+    pattern = re.compile(r"```json\s*(.*?)\s*```", re.DOTALL)
+
+    match = pattern.search(result)
+    if not match:
+        return np.empty((0, 4)), None, np.empty((0,), dtype=str)
+
+    json_snippet = match.group(1)
+
+    try:
+        data = json.loads(json_snippet)
+    except json.JSONDecodeError:
+        return np.empty((0, 4)), None, np.empty((0,), dtype=str)
+
+    boxes_list = []
+    labels_list = []
+
+    for item in data:
+        if "bbox_2d" not in item or "label" not in item:
+            continue
+        boxes_list.append(item["bbox_2d"])
+        labels_list.append(item["label"])
+
+    if not boxes_list:
+        return np.empty((0, 4)), None, np.empty((0,), dtype=str)
+
+    xyxy = np.array(boxes_list, dtype=float)
+    class_name = np.array(labels_list, dtype=str)
+
+    xyxy = xyxy / [in_w, in_h, in_w, in_h]
+    xyxy = xyxy * [out_w, out_h, out_w, out_h]
+
+    class_id = None
+
+    if classes is not None:
+        mask = np.array([label in classes for label in class_name], dtype=bool)
+        xyxy = xyxy[mask]
+        class_name = class_name[mask]
+        class_id = np.array([classes.index(label) for label in class_name], dtype=int)
+
+    return xyxy, class_id, class_name
+
+
 def from_florence_2(
     result: dict, resolution_wh: Tuple[int, int]
 ) -> Tuple[

supervision/keypoint/core.py

@@ -61,6 +61,7 @@ class KeyPoints:
         method, which accepts [MediaPipe](https://github.com/google-ai-edge/mediapipe)
         pose result.
 
+
         ```python
         import cv2
         import mediapipe as mp
@@ -91,10 +92,10 @@ class KeyPoints:
         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, m)` containing the confidence scores of each keypoint.
         class_id (Optional[np.ndarray]): An array of shape
             `(n,)` containing the class ids of the detected objects.
+        confidence (Optional[np.ndarray]): An array of shape
+            `(n, m)` containing the confidence scores of each keypoint.
         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 of length `n`
@@ -314,6 +315,7 @@ class KeyPoints:
             key_points = sv.KeyPoints.from_mediapipe(
                 face_landmarker_result, (image_width, image_height))
             ```
+
         """  # noqa: E501 // docs
         if hasattr(mediapipe_results, "pose_landmarks"):
             results = mediapipe_results.pose_landmarks
@@ -473,7 +475,7 @@ class KeyPoints:
             A `sv.KeyPoints` object containing the keypoint coordinates, class IDs,
                 and class names, and confidences of each keypoint.
 
-        Example:
+        Examples:
             ```python
             import cv2
             import supervision as sv
@@ -510,6 +512,91 @@ class KeyPoints:
         else:
             return cls.empty()
 
+    @classmethod
+    def from_transformers(cls, transfomers_results: Any) -> KeyPoints:
+        """
+        Create a `sv.KeyPoints` object from the
+        [Transformers](https://github.com/huggingface/transformers) inference result.
+
+        Args:
+            transfomers_results (Any): The output of a
+                Transformers model containing instances with prediction data.
+
+        Returns:
+            A `sv.KeyPoints` object containing the keypoint coordinates, class IDs,
+                and class names, and confidences of each keypoint.
+
+        Examples:
+            ```python
+            from PIL import Image
+            import requests
+            import supervision as sv
+            import torch
+            from transformers import (
+                AutoProcessor,
+                RTDetrForObjectDetection,
+                VitPoseForPoseEstimation,
+            )
+
+            device = "cuda" if torch.cuda.is_available() else "cpu"
+            image = Image.open(<SOURCE_IMAGE_PATH>)
+
+            DETECTION_MODEL_ID = "PekingU/rtdetr_r50vd_coco_o365"
+
+            detection_processor = AutoProcessor.from_pretrained(DETECTION_MODEL_ID, use_fast=True)
+            detection_model = RTDetrForObjectDetection.from_pretrained(DETECTION_MODEL_ID, device_map=DEVICE)
+
+            inputs = detection_processor(images=frame, return_tensors="pt").to(DEVICE)
+
+            with torch.no_grad():
+                outputs = detection_model(**inputs)
+
+            target_size = torch.tensor([(frame.height, frame.width)])
+            results = detection_processor.post_process_object_detection(
+                outputs, target_sizes=target_size, threshold=0.3)
+
+            detections = sv.Detections.from_transformers(results[0])
+            boxes = sv.xyxy_to_xywh(detections[detections.class_id == 0].xyxy)
+
+            POSE_ESTIMATION_MODEL_ID = "usyd-community/vitpose-base-simple"
+
+            pose_estimation_processor = AutoProcessor.from_pretrained(POSE_ESTIMATION_MODEL_ID)
+            pose_estimation_model = VitPoseForPoseEstimation.from_pretrained(
+                POSE_ESTIMATION_MODEL_ID, device_map=DEVICE)
+
+            inputs = pose_estimation_processor(frame, boxes=[boxes], return_tensors="pt").to(DEVICE)
+
+            with torch.no_grad():
+                outputs = pose_estimation_model(**inputs)
+
+            results = pose_estimation_processor.post_process_pose_estimation(outputs, boxes=[boxes])
+            key_point = sv.KeyPoints.from_transformers(results[0])
+            ```
+
+        """  # noqa: E501 // docs
+
+        if "keypoints" in transfomers_results[0]:
+            if transfomers_results[0]["keypoints"].cpu().numpy().size == 0:
+                return cls.empty()
+
+            result_data = [
+                (
+                    result["keypoints"].cpu().numpy(),
+                    result["scores"].cpu().numpy(),
+                )
+                for result in transfomers_results
+            ]
+
+            xy, scores = zip(*result_data)
+
+            return cls(
+                xy=np.stack(xy).astype(np.float32),
+                confidence=np.stack(scores).astype(np.float32),
+                class_id=np.arange(len(xy)).astype(int),
+            )
+        else:
+            return cls.empty()
+
     def __getitem__(
         self, index: Union[int, slice, List[int], np.ndarray, str]
     ) -> Union[KeyPoints, List, np.ndarray, None]:
@@ -639,7 +726,7 @@ class KeyPoints:
         Returns:
             detections (Detections): The converted detections object.
 
-        Example:
+        Examples:
             ```python
             keypoints = sv.KeyPoints.from_inference(...)
             detections = keypoints.as_detections()

supervision/metrics/utils/utils.py

@@ -5,5 +5,6 @@ def ensure_pandas_installed():
         raise ImportError(
             "`metrics` extra is required to run the function."
             " Run `pip install 'supervision[metrics]'` or"
-            " `poetry add supervision -E metrics`"
+            " `poetry add supervision -E metrics` or"
+            " `uv pip install 'supervision[metrics]'`"
         )

test/detection/test_lmm.py

@@ -1,151 +0,0 @@
-from typing import List, Optional, Tuple
-
-import numpy as np
-import pytest
-
-from supervision.detection.lmm import from_paligemma
-
-
-@pytest.mark.parametrize(
-    "result, resolution_wh, classes, expected_results",
-    [
-        (
-            "",
-            (1000, 1000),
-            None,
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # empty response
-        (
-            "",
-            (1000, 1000),
-            ["cat", "dog"],
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # empty response with classes
-        (
-            "\n",
-            (1000, 1000),
-            None,
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # new line response
-        (
-            "the quick brown fox jumps over the lazy dog.",
-            (1000, 1000),
-            None,
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # response with no location
-        (
-            "<loc0256><loc0768><loc0768> cat",
-            (1000, 1000),
-            None,
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # response with missing location
-        (
-            "<loc0256><loc0256><loc0768><loc0768><loc0768> cat",
-            (1000, 1000),
-            None,
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # response with extra location
-        (
-            "<loc0256><loc0256><loc0768><loc0768>",
-            (1000, 1000),
-            None,
-            (np.empty((0, 4)), None, np.empty(0).astype(str)),
-        ),  # response with no class
-        (
-            "<loc0256><loc0256><loc0768><loc0768> catt",
-            (1000, 1000),
-            ["cat", "dog"],
-            (np.empty((0, 4)), np.empty(0), np.empty(0).astype(str)),
-        ),  # response with invalid class
-        (
-            "<loc0256><loc0256><loc0768><loc0768> cat",
-            (1000, 1000),
-            None,
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                None,
-                np.array(["cat"]).astype(str),
-            ),
-        ),  # correct response; no classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> black cat",
-            (1000, 1000),
-            None,
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                None,
-                np.array(["black cat"]).astype(np.dtype("U")),
-            ),
-        ),  # correct response; class name with space; no classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> black-cat",
-            (1000, 1000),
-            None,
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                None,
-                np.array(["black-cat"]).astype(np.dtype("U")),
-            ),
-        ),  # correct response; class name with hyphen; no classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> black_cat",
-            (1000, 1000),
-            None,
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                None,
-                np.array(["black_cat"]).astype(np.dtype("U")),
-            ),
-        ),  # correct response; class name with underscore; no classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> cat ;",
-            (1000, 1000),
-            ["cat", "dog"],
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                np.array([0]),
-                np.array(["cat"]).astype(str),
-            ),
-        ),  # correct response; with classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> cat ; <loc0256><loc0256><loc0768><loc0768> dog",  # noqa: E501
-            (1000, 1000),
-            ["cat", "dog"],
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0], [250.0, 250.0, 750.0, 750.0]]),
-                np.array([0, 1]),
-                np.array(["cat", "dog"]).astype(np.dtype("U")),
-            ),
-        ),  # correct response; with classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> cat ; <loc0256><loc0256><loc0768> cat",  # noqa: E501
-            (1000, 1000),
-            ["cat", "dog"],
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                np.array([0]),
-                np.array(["cat"]).astype(str),
-            ),
-        ),  # partially correct response; with classes
-        (
-            "<loc0256><loc0256><loc0768><loc0768> cat ; <loc0256><loc0256><loc0768><loc0768><loc0768> cat",  # noqa: E501
-            (1000, 1000),
-            ["cat", "dog"],
-            (
-                np.array([[250.0, 250.0, 750.0, 750.0]]),
-                np.array([0]),
-                np.array(["cat"]).astype(str),
-            ),
-        ),  # partially correct response; with classes
-    ],
-)
-def test_from_paligemma(
-    result: str,
-    resolution_wh: Tuple[int, int],
-    classes: Optional[List[str]],
-    expected_results: Tuple[np.ndarray, Optional[np.ndarray], np.ndarray],
-) -> None:
-    result = from_paligemma(result=result, resolution_wh=resolution_wh, classes=classes)
-    np.testing.assert_array_equal(result[0], expected_results[0])
-    np.testing.assert_array_equal(result[1], expected_results[1])
-    np.testing.assert_array_equal(result[2], expected_results[2])

test/detection/test_utils.py

@@ -21,6 +21,7 @@ from supervision.detection.utils import (
     scale_boxes,
     xcycwh_to_xyxy,
     xywh_to_xyxy,
+    xyxy_to_xywh,
 )
 
 TEST_MASK = np.zeros((1, 1000, 1000), dtype=bool)
@@ -1381,6 +1382,29 @@ def test_xywh_to_xyxy(xywh: np.ndarray, expected_result: np.ndarray) -> None:
     np.testing.assert_array_equal(result, expected_result)
 
 
+@pytest.mark.parametrize(
+    "xyxy, expected_result",
+    [
+        (np.array([[10, 20, 40, 60]]), np.array([[10, 20, 30, 40]])),  # standard case
+        (np.array([[0, 0, 0, 0]]), np.array([[0, 0, 0, 0]])),  # zero size bounding box
+        (
+            np.array([[50, 50, 150, 150]]),
+            np.array([[50, 50, 100, 100]]),
+        ),  # large bounding box
+        (
+            np.array([[-10, -20, 20, 20]]),
+            np.array([[-10, -20, 30, 40]]),
+        ),  # negative coordinates
+        (np.array([[50, 50, 50, 80]]), np.array([[50, 50, 0, 30]])),  # zero width
+        (np.array([[50, 50, 70, 50]]), np.array([[50, 50, 20, 0]])),  # zero height
+        (np.array([]).reshape(0, 4), np.array([]).reshape(0, 4)),  # empty array
+    ],
+)
+def test_xyxy_to_xywh(xyxy: np.ndarray, expected_result: np.ndarray) -> None:
+    result = xyxy_to_xywh(xyxy)
+    np.testing.assert_array_equal(result, expected_result)
+
+
 @pytest.mark.parametrize(
     "xcycwh, expected_result",
     [

test/detection/test_vlm.py

@@ -0,0 +1,355 @@
+from contextlib import nullcontext as does_not_raise
+from typing import List, Optional, Tuple
+
+import numpy as np
+import pytest
+
+from supervision.detection.vlm import from_paligemma, from_qwen_2_5_vl
+
+
+@pytest.mark.parametrize(
+    "exception, result, resolution_wh, classes, expected_results",
+    [
+        (
+            does_not_raise(),
+            "",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # empty text
+        (
+            does_not_raise(),
+            "",
+            (1000, 1000),
+            ["cat", "dog"],
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # empty text, classes
+        (
+            does_not_raise(),
+            "\n",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # newline only
+        (
+            does_not_raise(),
+            "the quick brown fox jumps over the lazy dog.",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # random text, no location
+        (
+            does_not_raise(),
+            "<loc0256><loc0768><loc0768> cat",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # partial location
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768><loc0768> cat",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # extra loc
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768>",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # no class
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> catt",
+            (1000, 1000),
+            ["cat", "dog"],
+            (np.empty((0, 4)), np.empty(0), np.empty(0).astype(str)),
+        ),  # invalid class
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # single box, no classes
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> black cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["black cat"]).astype(np.dtype("U")),
+            ),
+        ),  # class with space
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> black-cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["black-cat"]).astype(np.dtype("U")),
+            ),
+        ),  # class with hyphen
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> black_cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["black_cat"]).astype(np.dtype("U")),
+            ),
+        ),  # class with underscore
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> cat ;",
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                np.array([0]),
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # correct class filter
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> cat ; "
+            "<loc0256><loc0256><loc0768><loc0768> dog",
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0], [250.0, 250.0, 750.0, 750.0]]),
+                np.array([0, 1]),
+                np.array(["cat", "dog"]).astype(np.dtype("U")),
+            ),
+        ),  # multiple correct boxes, classes
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> cat ; "
+            "<loc0256><loc0256><loc0768> cat",
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                np.array([0]),
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # partial valid boxes
+        (
+            does_not_raise(),
+            "<loc0256><loc0256><loc0768><loc0768> cat ; "
+            "<loc0256><loc0256><loc0768><loc0768><loc0768> cat",
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                np.array([0]),
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # partial valid again
+        (
+            pytest.raises(ValueError),
+            "<loc0256><loc0256><loc0768><loc0768> cat",
+            (0, 1000),
+            None,
+            None,
+        ),  # zero width -> ValueError
+        (
+            pytest.raises(ValueError),
+            "<loc0256><loc0256><loc0768><loc0768> dog",
+            (1000, -200),
+            None,
+            None,
+        ),  # negative height -> ValueError
+    ],
+)
+def test_from_paligemma(
+    exception,
+    result: str,
+    resolution_wh: Tuple[int, int],
+    classes: Optional[List[str]],
+    expected_results: Tuple[np.ndarray, Optional[np.ndarray], np.ndarray],
+) -> None:
+    with exception:
+        result = from_paligemma(
+            result=result, resolution_wh=resolution_wh, classes=classes
+        )
+        np.testing.assert_array_equal(result[0], expected_results[0])
+        np.testing.assert_array_equal(result[1], expected_results[1])
+        np.testing.assert_array_equal(result[2], expected_results[2])
+
+
+@pytest.mark.parametrize(
+    "exception, result, input_wh, resolution_wh, classes, expected_results",
+    [
+        (
+            does_not_raise(),
+            "some random text without triple backticks",
+            (640, 640),
+            (1280, 720),
+            None,
+            (np.empty((0, 4)), None, np.empty(0, dtype=str)),
+        ),  # no snippet
+        (
+            does_not_raise(),
+            "```json\nnot valid json\n```",
+            (640, 640),
+            (1280, 720),
+            None,
+            (np.empty((0, 4)), None, np.empty(0, dtype=str)),
+        ),  # invalid JSON
+        (
+            does_not_raise(),
+            "```json\n[]\n```",
+            (640, 640),
+            (1280, 720),
+            None,
+            (np.empty((0, 4)), None, np.empty(0, dtype=str)),
+        ),  # empty list
+        (
+            does_not_raise(),
+            """```json
+            [
+                {"bbox_2d": [10, 10, 100, 100]},
+                {"label": "missing box"},
+                {"bbox_2d": [50, 60, 110, 120], "unused": "something"}
+            ]
+            ```""",
+            (640, 640),
+            (1280, 720),
+            None,
+            (np.empty((0, 4)), None, np.empty(0, dtype=str)),
+        ),  # missing keys
+        (
+            does_not_raise(),
+            """```json
+            [
+                {"bbox_2d": [10, 20, 110, 120], "label": "cat"}
+            ]
+            ```""",
+            (640, 640),
+            (1280, 720),
+            None,
+            (
+                np.array([[20.0, 22.5, 220.0, 135.0]]),
+                None,
+                np.array(["cat"], dtype=str),
+            ),
+        ),  # single box no classes
+        (
+            does_not_raise(),
+            """```json
+            [
+                {"bbox_2d": [0, 0, 64, 64], "label": "dog"},
+                {"bbox_2d": [100, 200, 300, 400], "label": "cat"}
+            ]
+            ```""",
+            (640, 640),
+            (640, 640),
+            None,
+            (
+                np.array([[0, 0, 64, 64], [100, 200, 300, 400]], dtype=float),
+                None,
+                np.array(["dog", "cat"], dtype=str),
+            ),
+        ),  # multiple no classes
+        (
+            does_not_raise(),
+            """```json
+            [
+                {"bbox_2d": [10, 20, 110, 120], "label": "bird"}
+            ]
+            ```""",
+            (640, 640),
+            (1280, 720),
+            ["cat", "dog"],
+            (np.empty((0, 4)), np.empty(0, dtype=int), np.empty(0, dtype=str)),
+        ),  # class mismatch
+        (
+            does_not_raise(),
+            """```json
+            [
+                {"bbox_2d": [10, 20, 110, 120], "label": "cat"},
+                {"bbox_2d": [50, 100, 150, 200], "label": "dog"}
+            ]
+            ```""",
+            (640, 640),
+            (640, 480),
+            ["cat", "dog"],
+            (
+                np.array([[10.0, 15.0, 110.0, 90.0], [50.0, 75.0, 150.0, 150.0]]),
+                np.array([0, 1], dtype=int),
+                np.array(["cat", "dog"], dtype=str),
+            ),
+        ),  # partial filtering
+        (
+            does_not_raise(),
+            """```json
+            [
+                {"bbox_2d": [-10, 0, 700, 700], "label": "dog"}
+            ]
+            ```""",
+            (640, 640),
+            (1280, 720),
+            None,
+            (
+                np.array([[-20.0, 0.0, 1400.0, 787.5]]),
+                None,
+                np.array(["dog"], dtype=str),
+            ),
+        ),  # out-of-bounds box
+        (
+            pytest.raises(ValueError),
+            """```json
+            [
+                {"bbox_2d": [10, 20, 110, 120], "label": "cat"}
+            ]
+            ```""",
+            (0, 640),
+            (1280, 720),
+            None,
+            None,  # won't be compared because we expect an exception
+        ),  # zero input width -> ValueError
+        (
+            pytest.raises(ValueError),
+            """```json
+            [
+                {"bbox_2d": [10, 20, 110, 120], "label": "dog"}
+            ]
+            ```""",
+            (640, 640),
+            (1280, -100),
+            None,
+            None,
+        ),  # negative resolution height -> ValueError
+    ],
+)
+def test_from_qwen_2_5_vl(
+    exception,
+    result: str,
+    input_wh: Tuple[int, int],
+    resolution_wh: Tuple[int, int],
+    classes: Optional[List[str]],
+    expected_results,
+) -> None:
+    with exception:
+        xyxy, class_id, class_name = from_qwen_2_5_vl(
+            result=result,
+            input_wh=input_wh,
+            resolution_wh=resolution_wh,
+            classes=classes,
+        )
+        if expected_results is not None:
+            np.testing.assert_array_equal(xyxy, expected_results[0])
+            np.testing.assert_array_equal(class_id, expected_results[1])
+            np.testing.assert_array_equal(class_name, expected_results[2])

test/detection/test_vlm_florence_2.py

@@ -4,7 +4,7 @@ from typing import Optional, Tuple
 import numpy as np
 import pytest
 
-from supervision.detection.lmm import from_florence_2
+from supervision.detection.vlm import from_florence_2
 
 
 @pytest.mark.parametrize(