bump version from 0.26.0rc6 to 0.26.0rc7

fix: `UnboundLocalError` in `process_video`

.github/CODEOWNERS

@@ -0,0 +1,4 @@
+# These owners will be the default owners for everything in
+# the repo. They will be requested for review when someone
+# opens a pull request.
+*   @SkalskiP @onuralpszr

.github/dependabot.yml

@@ -4,13 +4,15 @@ updates:
   - package-ecosystem: "github-actions"
     directory: "/"
     schedule:
-      interval: "daily"
+      interval: "weekly"
     commit-message:
       prefix: ⬆️
+    target-branch: "develop"
   # Python
   - package-ecosystem: "pip"
     directory: "/"
     schedule:
-      interval: "daily"
+      interval: "weekly"
     commit-message:
       prefix: ⬆️
+    target-branch: "develop"

.github/workflows/clear-cache.yml

@@ -1,33 +1,42 @@
-
 name: Clear cache
 
 on:
-    schedule:
-      - cron: '0 0 1 * *'
-    workflow_dispatch:
+  schedule:
+    - cron: "0 0 1 * *" # Run at midnight on the first day of every month
+  workflow_dispatch:
 
+# Restrict permissions by default
 permissions:
-  actions: write
+  actions: write # Required for cache management
 
 jobs:
   clear-cache:
+    name: Clear cache
     runs-on: ubuntu-latest
+    timeout-minutes: 10
     steps:
       - name: Clear cache
-        uses: actions/github-script@v7
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
         with:
           script: |
-            console.log("About to clear")
+            console.log("Starting cache cleanup...")
             const caches = await github.rest.actions.getActionsCacheList({
               owner: context.repo.owner,
               repo: context.repo.repo,
             })
+
+            let deletedCount = 0
             for (const cache of caches.data.actions_caches) {
-              console.log(cache)
-              github.rest.actions.deleteActionsCacheById({
-                owner: context.repo.owner,
-                repo: context.repo.repo,
-                cache_id: cache.id,
-              })
+              console.log(`Deleting cache: ${cache.key} (${cache.size_in_bytes} bytes)`)
+              try {
+                await github.rest.actions.deleteActionsCacheById({
+                  owner: context.repo.owner,
+                  repo: context.repo.repo,
+                  cache_id: cache.id,
+                })
+                deletedCount++
+              } catch (error) {
+                console.error(`Failed to delete cache ${cache.key}: ${error.message}`)
+              }
             }
-            console.log("Clear completed")
+            console.log(`Cache cleanup completed. Deleted ${deletedCount} caches.`)

.github/workflows/combine-dependabot-prs.yml

@@ -0,0 +1,22 @@
+name: Combine Dependabot PRs
+
+on:
+  schedule:
+    - cron: "0 1 * * 3" # Wednesday at 01:00
+  workflow_dispatch: # allows you to manually trigger the workflow
+
+permissions:
+  contents: write
+  pull-requests: write
+  checks: read
+
+jobs:
+  combine-prs:
+    name: Combine
+    runs-on: ubuntu-latest
+    steps:
+      - name: combine-prs
+        id: combine-prs
+        uses: github/combine-prs@2909f404763c3177a456e052bdb7f2e85d3a7cb3 # v5.2.0
+        with:
+          labels: combined-pr

.github/workflows/notebook-bot.yml

@@ -1,64 +0,0 @@
-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/poetry-test.yml

@@ -1,37 +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:
-    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@v4
-
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: 📦 Install the base dependencies
-        run: python -m pip install --upgrade poetry
-
-      - name: 🔍 Check the correctness of the project config
-        run: poetry check
-
-      - name: 🚀 Do Install the package Test
-        run: poetry install

.github/workflows/publish-dev-docs.yml

@@ -1,53 +0,0 @@
-name: Docs WorkFlow - Develop Tag 📚
-
-on:
-  push:
-    branches:
-      - develop
-  workflow_dispatch:
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event_name == 'push' && github.ref}}
-  cancel-in-progress: true
-
-permissions:
-  contents: write
-  pages: write
-  pull-requests: write
-
-
-jobs:
-  deploy:
-    runs-on: ubuntu-latest
-    steps:
-      - name: 🔄 Checkout code
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-      - name: 🐍 Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.10'
-      - name: 📦 Install mkdocs-material
-        run: pip install "mkdocs-material"
-      - 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: ⚙️ Configure git for github-actions
-        run: |
-          git config --global user.name "github-actions[bot]"
-          git config --global user.email "41898282+github-actions[bot]@users.noreply.github.com"
-      - name: 🚀 Deploy MkDoc-Material with mike
-        run: |
-          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} mike deploy --push develop

.github/workflows/publish-docs.yml

@@ -0,0 +1,70 @@
+name: Build and Publish Docs
+
+on:
+  push:
+    branches:
+      - develop
+  workflow_dispatch:
+  release:
+    types: [published]
+
+# Ensure only one concurrent deployment
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event_name == 'push' && github.ref}}
+  cancel-in-progress: true
+
+# Restrict permissions by default
+permissions:
+  contents: write # Required for committing to gh-pages
+  pages: write # Required for deploying to Pages
+  pull-requests: write # Required for PR comments
+
+jobs:
+  deploy:
+    name: Publish Docs
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    strategy:
+      matrix:
+        python-version: ["3.10"]
+    steps:
+      - name: 📥 Checkout the repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+
+      - name: 🐍 Install uv and set Python ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: 🔑 Create GitHub App token (mkdocs)
+        id: mkdocs_token
+        uses: actions/create-github-app-token@3ff1caaa28b64c9cc276ce0a02e2ff584f3900c5 # v2.0.2
+        with:
+          app-id: ${{ secrets.MKDOCS_APP_ID }}
+          private-key: ${{ secrets.MKDOCS_PEM }}
+          owner: roboflow
+          repositories: mkdocs-material-insiders
+
+      - name: 🏗️ Install dependencies
+        run: |
+          uv pip install -r pyproject.toml --extra docs
+          # Install mkdocs-material-insiders using the GitHub App token
+          uv pip install "git+https://roboflow:${{ 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: |
+          git config --global user.name "github-actions[bot]"
+          git config --global user.email "41898282+github-actions[bot]@users.noreply.github.com"
+
+      - name: 🚀 Deploy Development Docs
+        if: (github.event_name == 'push' && github.ref == 'refs/heads/develop') || github.event_name == 'workflow_dispatch'
+        run: |
+          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} uv run mike deploy --push develop
+
+      - name: 🚀 Deploy Release Docs
+        if: github.event_name == 'release' && github.event.action == 'published'
+        run: |
+          latest_tag=$(git describe --tags `git rev-list --tags --max-count=1`)
+          MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} 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@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
+        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-docs.yml

@@ -1,55 +0,0 @@
-name: Supervision Release Documentation Workflow 📚
-on:
-  workflow_dispatch:
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event_name == 'push' && github.ref}}
-  cancel-in-progress: true
-
-permissions:
-  contents: write
-  pages: write
-  pull-requests: write
-
-
-jobs:
-  doc-build-deploy:
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: ["3.10"]
-    steps:
-      - name: 🛎️ Checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          ref: ${{ github.head_ref }}
-
-      - name: 🐍 Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.10'
-      - name: 📦 Install mkdocs-material
-        run: pip install "mkdocs-material"
-      - name: 📦 Install mkdocstrings[python]
-        run: pip install "mkdocstrings[python]"
-      - name: 📦 Install mkdocs-material[imaging]
-        run: pip install "mkdocs-material[imaging]"
-      - name: 📦 Install mike
-        run: pip install "mike"
-      - name: 📦 Install mkdocs-git-revision-date-localized-plugin
-        run: pip install "mkdocs-git-revision-date-localized-plugin"
-      - name: 📦 Install JupyterLab
-        run: pip install jupyterlab
-      - name: 📦 Install mkdocs-jupyter
-        run: pip install mkdocs-jupyter
-      - name: 📦 Install mkdocs-git-committers-plugin-2
-        run: pip install mkdocs-git-committers-plugin-2
-      - name: ⚙️ Configure git for github-actions 👷
-        run: |
-          git config --global user.name "github-actions[bot]"
-          git config --global user.email "41898282+github-actions[bot]@users.noreply.github.com"
-      - name: 🚀 Deploy MkDoc-Material 📚
-        run: |
-            latest_tag=$(git describe --tags `git rev-list --tags --max-count=1`)
-            MKDOCS_GIT_COMMITTERS_APIKEY=${{ secrets.GITHUB_TOKEN }} mike deploy --push --update-aliases $latest_tag latest

.github/workflows/publish-release.yml

@@ -0,0 +1,43 @@
+name: Publish Supervision Releases to PyPI
+
+on:
+  push:
+    tags:
+      - "[0-9]+.[0-9]+[0-9]+.[0-9]"
+  workflow_dispatch:
+
+permissions: {} # Explicitly remove all permissions by default
+
+jobs:
+  publish-release:
+    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@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
+        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-test.yml

@@ -1,43 +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:
-
-jobs:
-  build-and-publish-pre-release-pypi:
-    name: Build and publish to PyPI
-    runs-on: ubuntu-latest
-    environment: test
-    permissions:
-      id-token: write
-    strategy:
-      matrix:
-        python-version: ["3.10"]
-    steps:
-      - name: 🛎️ Checkout
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ github.head_ref }}
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: 🏗️ Build source and wheel distributions
-        run: |
-          python -m pip install --upgrade build twine
-          python -m build
-          twine check --strict dist/*
-
-      - name: 🚀 Publish to PyPi
-        uses: pypa/gh-action-pypi-publish@release/v1.10
-
-      - name: 🚀 Publish to Test-PyPi
-        uses: pypa/gh-action-pypi-publish@release/v1.10
-        with:
-          repository-url: https://test.pypi.org/legacy/

.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@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
+        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/publish.yml

@@ -1,40 +0,0 @@
-name: Publish Supervision Releases to PyPI and TestPyPI
-on:
-  push:
-    tags:
-      - "[0-9]+.[0-9]+[0-9]+.[0-9]"
-
-  workflow_dispatch:
-
-jobs:
-  build-and-publish-pre-release:
-    runs-on: ubuntu-latest
-    environment: release
-    permissions:
-      id-token: write
-    strategy:
-      matrix:
-        python-version: ["3.10"]
-    steps:
-      - name: 🛎️ Checkout
-        uses: actions/checkout@v4
-        with:
-          ref: ${{ github.head_ref }}
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-
-      - name: 🏗️ Build source and wheel distributions
-        run: |
-          python -m pip install --upgrade build twine
-          python -m build
-          twine check --strict dist/*
-
-      - name: 🚀 Publish to PyPi
-        uses: pypa/gh-action-pypi-publish@release/v1.10
-
-      - name: 🚀 Publish to Test-PyPi
-        uses: pypa/gh-action-pypi-publish@release/v1.10
-        with:
-          repository-url: https://test.pypi.org/legacy/

.github/workflows/test-doc.yml

@@ -4,18 +4,34 @@ on:
   pull_request:
     branches: [main, develop]
 
+# Restrict permissions by default
+permissions:
+  contents: read # Required for checkout
+  checks: write # Required for test reporting
+
 jobs:
   docs-build-test:
+    name: Test docs build
     runs-on: ubuntu-latest
+    timeout-minutes: 10
+    strategy:
+      matrix:
+        python-version: ["3.10"]
     steps:
-      - name: 🔄 Checkout code
-        uses: actions/checkout@v4
-      - name: 🐍 Set up Python
-        uses: actions/setup-python@v5
+      - name: 📥 Checkout the repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
         with:
-          python-version: '3.10'
-      - name:  🏗️ Install dependencies and Test Docs Build
+          fetch-depth: 0
+
+      - name: 🐍 Install uv and set Python ${{ matrix.python-version }}
+        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: 🏗️ Install dependencies
         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
+          uv pip install -r pyproject.toml --extra docs
+
+
+      - name: 🧪 Test Docs Build
+        run: uv run mkdocs build --verbose

.github/workflows/test-min.yml

@@ -1,58 +0,0 @@
-name: Python 3.8 - Min Dep Test WorkFlow
-
-on:
-  pull_request:
-    branches: [main, develop]
-
-jobs:
-  build-min-dep-test:
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: ["3.8"]
-    steps:
-      - name: 🛎️ Checkout
-        uses: actions/checkout@v4
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        # id based on python version
-        id: python-setup
-        with:
-          python-version: ${{ matrix.python-version }}
-          check-latest: true
-
-      - name: 📦 Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install \
-              attrs==23.1.0 \
-              certifi==2023.7.22 \
-              charset-normalizer==2.0.12 \
-              cycler==0.12.1 \
-              exceptiongroup==1.1.3 \
-              fonttools==4.43.1 \
-              idna==3.4 \
-              iniconfig==2.0.0 \
-              kiwisolver==1.4.5 \
-              matplotlib==3.5.0 \
-              numpy==1.21.2 \
-              opencv-python==4.5.5.64 \
-              Pillow==10.1.0 \
-              packaging==23.2 \
-              pluggy==1.3.0 \
-              pyparsing==3.1.1 \
-              pytest==7.2.0 \
-              python-dateutil==2.8.2 \
-              PyYAML==5.3 \
-              requests==2.26.0 \
-              scipy==1.10.0 \
-              setuptools-scm==8.0.4 \
-              six==1.16.0 \
-              tomli==2.0.1 \
-              tqdm==4.62.3 \
-              typing_extensions==4.8.0 \
-              urllib3==1.26.18 \
-              defusedxml==0.7.1
-
-      - name: 🧪 Test
-        run: "python -m pytest ./test"

.github/workflows/test.yml

@@ -1,29 +0,0 @@
-name: Test WorkFlow
-
-on:
-  pull_request:
-    branches: [main, develop]
-
-jobs:
-  build-dev-test:
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: ["3.8", "3.9", "3.10", "3.11", "3.12", "3.13"]
-    steps:
-      - name: 🛎️ Checkout
-        uses: actions/checkout@v4
-      - name: 🐍 Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python-version }}
-          check-latest: true
-
-      - name: 📦 Install dependencies
-        run: |
-          python -m pip install --upgrade pip
-          pip install .
-          pip install pytest
-
-      - name: 🧪 Test
-        run: "python -m pytest ./test"

.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@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5.4.2
+        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

@@ -25,14 +25,14 @@ repos:
       - id: mixed-line-ending
 
   -   repo: https://github.com/PyCQA/bandit
-      rev: '1.7.10'
+      rev: '1.8.3'
       hooks:
       -   id: bandit
           args: ["-c", "pyproject.toml"]
           additional_dependencies: ["bandit[toml]"]
 
   -   repo: https://github.com/astral-sh/ruff-pre-commit
-      rev: v0.7.3
+      rev: v0.11.6
       hooks:
       -   id: ruff
           args: [--fix, --exit-non-zero-on-fix]
@@ -48,7 +48,7 @@ repos:
   #       args: ["--number"]
 
   - repo: https://github.com/codespell-project/codespell
-    rev: v2.3.0
+    rev: v2.4.1
     hooks:
       - id: codespell
         additional_dependencies:

CONTRIBUTING.md

@@ -128,15 +128,26 @@ PRs must pass all tests and linting requirements before they can be merged.
 
 Before starting your work on the project, set up your development environment:
 
-1. Clone your fork of the project:
+1. Clone your fork of the project (recommended to use shallow clone of develop branch):
+
+    **Option A: Recommended for most contributors (shallow clone of develop branch):**
+
+    ```bash
+    git clone --depth 1 -b develop https://github.com/YOUR_USERNAME/supervision.git
+    cd supervision
+    ```
+
+    Replace `YOUR_USERNAME` with your GitHub username.
+
+    > Note: Using `--depth 1` creates a shallow clone with minimal history and `-b develop` ensures you start with the development branch. This significantly reduces download size while providing everything needed to contribute.
+
+    **Option B: Full repository clone (if you need complete history):**
 
     ```bash
     git clone https://github.com/YOUR_USERNAME/supervision.git
     cd supervision
     ```
 
-    Replace `YOUR_USERNAME` with your GitHub username.
-
 2. Create and activate a virtual environment:
 
     ```bash
@@ -144,31 +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
@@ -181,7 +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.
 
@@ -203,7 +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/).
 

demo.ipynb

@@ -353,7 +353,7 @@
    },
    "outputs": [],
    "source": [
-    "!pip install -q ultralytics"
+    "!pip install -q \"ultralytics<=8.3.40\""
    ]
   },
   {

docs/assets.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Assets
@@ -8,17 +7,6 @@ status: new
 Supervision offers an assets download utility that allows you to download video files
 that you can use in your demos.
 
-## Install extra
-
-To install the Supervision assets utility, you can use `pip`. This utility is available
-as an extra within the Supervision package.
-
-!!! example "pip install"
-
-    ```bash
-    pip install "supervision[assets]"
-    ```
-
 <div class="md-typeset">
     <h2><a href="#supervision.assets.downloader.download_assets.download_assets">download_assets</a></h2>
 </div>

docs/detection/annotators.md

@@ -28,26 +28,28 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     </div>
 
-=== "RoundBox"
+    <!--
+    === "RoundBox"
 
-    ```python
-    import supervision as sv
+        ```python
+        import supervision as sv
 
-    image = ...
-    detections = sv.Detections(...)
+        image = ...
+        detections = sv.Detections(...)
 
-    round_box_annotator = sv.RoundBoxAnnotator()
-    annotated_frame = round_box_annotator.annotate(
-        scene=image.copy(),
-        detections=detections
-    )
-    ```
+        round_box_annotator = sv.RoundBoxAnnotator()
+        annotated_frame = round_box_annotator.annotate(
+            scene=image.copy(),
+            detections=detections
+        )
+        ```
 
-    <div class="result" markdown>
+        <div class="result" markdown>
 
-    ![round-box-annotator-example](https://media.roboflow.com/supervision-annotator-examples/round-box-annotator-example-purple.png){ align=center width="800" }
+        ![round-box-annotator-example](https://media.roboflow.com/supervision-annotator-examples/round-box-annotator-example-purple.png){ align=center width="800" }
 
-    </div>
+        </div>
+    -->
 
 === "BoxCorner"
 
@@ -234,7 +236,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![mask-annotator-example](https://media.roboflow.com/supervision-annotator-examples/mask-annotator-example-purple.png){ align=center width="800" }
+    ![mask-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    mask-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -255,7 +258,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![polygon-annotator-example](https://media.roboflow.com/supervision-annotator-examples/polygon-annotator-example-purple.png){ align=center width="800" }
+    ![polygon-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    polygon-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -283,7 +287,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/label-annotator-example-purple.png){ align=center width="800" }
+    ![label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    label-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -314,7 +319,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/label-annotator-example-purple.png){ align=center width="800" }
+    ![label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    label-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -341,24 +347,32 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![icon-annotator-example](https://media.roboflow.com/supervision-annotator-examples/icon-annotator-example.png){ align=center width="800" }
+    ![icon-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    icon-annotator-example.png){ align=center width="800" }
 
     </div>
 
-=== "Crop"
+    <!-- === "Crop"
 
-    ```python
-    import supervision as sv
+        ```python
+        import supervision as sv
 
-    image = ...
-    detections = sv.Detections(...)
+        image = ...
+        detections = sv.Detections(...)
 
-    crop_annotator = sv.CropAnnotator()
-    annotated_frame = crop_annotator.annotate(
-        scene=image.copy(),
-        detections=detections
-    )
-    ```
+        crop_annotator = sv.CropAnnotator()
+        annotated_frame = crop_annotator.annotate(
+            scene=image.copy(),
+            detections=detections
+        )
+        ```
+        <div class="result" markdown>
+
+        ![crop-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+        crop-annotator-example.png){ align=center width="800" }
+
+        </div>
+    -->
 
 === "Blur"
 
@@ -377,7 +391,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![blur-annotator-example](https://media.roboflow.com/supervision-annotator-examples/blur-annotator-example-purple.png){ align=center width="800" }
+    ![blur-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    blur-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -398,7 +413,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![pixelate-annotator-example](https://media.roboflow.com/supervision-annotator-examples/pixelate-annotator-example-10.png){ align=center width="800" }
+    ![pixelate-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    pixelate-annotator-example-10.png){ align=center width="800" }
 
     </div>
 
@@ -429,7 +445,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![trace-annotator-example](https://media.roboflow.com/supervision-annotator-examples/trace-annotator-example-purple.png){ align=center width="800" }
+    ![trace-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    trace-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -458,7 +475,8 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![heat-map-annotator-example](https://media.roboflow.com/supervision-annotator-examples/heat-map-annotator-example-purple.png){ align=center width="800" }
+    ![heat-map-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    heat-map-annotator-example-purple.png){ align=center width="800" }
 
     </div>
 
@@ -479,7 +497,31 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
     <div class="result" markdown>
 
-    ![background-overlay-annotator-example](https://media.roboflow.com/supervision-annotator-examples/background-color-annotator-example-purple.png)
+    ![background-overlay-annotator-example](https://media.roboflow.com/supervision-annotator-examples/background-color-annotator-example-purple.png){ align=center width="800" }
+
+    </div>
+
+=== "Comparison"
+
+    ```python
+    import supervision as sv
+
+    image = ...
+    detections_1 = sv.Detections(...)
+    detections_2 = sv.Detections(...)
+
+    comparison_annotator = sv.ComparisonAnnotator()
+    annotated_frame = comparison_annotator.annotate(
+        scene=image.copy(),
+        detections_1=detections_1,
+        detections_2=detections_2
+    )
+    ```
+
+    <div class="result" markdown>
+
+    ![comparison-annotator-example](https://media.roboflow.com/supervision-annotator-examples/
+    comparison-annotator-example.png){ align=center width="800" }
 
     </div>
 
@@ -622,6 +664,12 @@ Annotators accept detections and apply box or mask visualizations to the detecti
 
 :::supervision.annotators.core.BackgroundOverlayAnnotator
 
+<div class="md-typeset">
+    <h2><a href="#supervision.annotators.core.ComparisonAnnotator">ComparisonAnnotator</a></h2>
+</div>
+
+:::supervision.annotators.core.ComparisonAnnotator
+
 <div class="md-typeset">
     <h2><a href="#supervision.annotators.core.ColorLookup">ColorLookup</a></h2>
 </div>

docs/detection/utils.md

@@ -89,6 +89,18 @@ 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.xyxy_to_xcycarh">xyxy_to_xcycarh</a></h2>
+</div>
+
+:::supervision.detection.utils.xyxy_to_xcycarh
+
 <div class="md-typeset">
   <h2><a href="#supervision.detection.utils.xcycwh_to_xyxy">xcycwh_to_xyxy</a></h2>
 </div>

docs/how_to/benchmark_a_model.md

@@ -0,0 +1,441 @@
+---
+comments: true
+status: new
+---
+
+![Corgi Example](https://media.roboflow.com/supervision/image-examples/how-to/benchmark-models/corgi-sorted-2.png)
+
+# Benchmark a Model
+
+Have you ever trained multiple detection models and wondered which one performs best on your specific use case? Or maybe you've downloaded a pre-trained model and want to verify its performance on your dataset? Model benchmarking is essential for making informed decisions about which model to deploy in production.
+
+This guide will show an easy way to benchmark your results using `supervision`. It will go over:
+
+1. [Loading a dataset](#loading-a-dataset)
+2. [Loading a model](#loading-a-model)
+3. [Benchmarking Basics](#benchmarking-basics)
+4. [Running a Model](#running-a-model)
+5. [Remapping Classes](#remapping-classes)
+6. [Visual Benchmarking](#visual-benchmarking)
+7. [Benchmarking Metrics](#benchmarking-metrics)
+8. [Mean Average Precision (mAP)](#mean-average-precision-map)
+9. [F1 Score](#f1-score)
+10. [Bonus: Model Leaderboard](#model-leaderboard)
+
+This guide will use an instance segmentation model, but it applies to object detection, instance segmentation, and oriented bounding box models (OBB) too.
+
+A condensed version of this guide is available as a [Colab Notebook](https://colab.research.google.com/drive/1HoOY9pZoVwGiRMmLHtir0qT6Uj45w6Ps?usp=sharing).
+
+## Loading a Dataset
+
+Suppose you start with a dataset. Perhaps you found it on [Universe](https://universe.roboflow.com/); perhaps you [labeled your own](https://roboflow.com/how-to-label/yolo11). In either case, this guide assumes you know of a labelled dataset at hand.
+
+We'll use the following libraries:
+
+- `roboflow` to manage the dataset and deploy models
+- `inference` to run the models
+- `supervision` to evaluate the model results
+
+```bash
+pip install roboflow supervision
+pip install git+https://github.com/roboflow/inference.git@linas/allow-latest-rc-supervision
+```
+
+!!! info
+
+    We're updating `inference` at the moment. Please install it as shown above.
+
+Here's how you can download a dataset:
+
+```python
+from roboflow import Roboflow
+
+rf = Roboflow(api_key="<YOUR_API_KEY>")
+project = rf.workspace("<WORKSPACE_NAME>").project("<PROJECT_NAME>")
+dataset = project.version(<DATASET_VERSION_NUMBER>).download("<FORMAT>")
+```
+
+If your dataset is from Universe, go to `Dataset` > `Download Dataset` > select the format (e.g. `YOLOv11`) > `Show download code`.
+
+If labeling your own data, go to the [dashboard](https://app.roboflow.com/) and check this [guide](https://docs.roboflow.com/api-reference/workspace-and-project-ids) to find your workspace and project IDs.
+
+In this guide, we shall use a small [Corgi v2](https://universe.roboflow.com/model-examples/segmented-animals-basic) dataset. It is well-labeled and comes with a test set.
+
+```python
+from roboflow import Roboflow
+
+rf = Roboflow(api_key="<YOUR_API_KEY>")
+project = rf.workspace("fbamse1-gm2os").project("corgi-v2")
+dataset = project.version(4).download("yolov11")
+```
+
+This will create a folder called `Corgi-v2-4` with the dataset in the current working directory, with `train`, `test`, and `valid` folders and a `data.yaml` file.
+
+## Loading a Model
+
+Let's load a model.
+
+=== "Inference, Local"
+
+    Roboflow supports a range of state-of-the-art [pre-trained models](https://inference.roboflow.com/quickstart/aliases/) for object detection, instance segmentation, and pose tracking. You don't even need an API key!
+
+    Let's load such a model with inference [`inference`](https://inference.roboflow.com/).
+
+    ```python
+    from inference import get_model
+
+    model = get_model(model_id="yolov11s-seg-640")
+    ```
+
+=== "Inference, Deployed"
+
+    You can train and deploy a model without leaving the Roboflow platform. See this [guide](https://docs.roboflow.com/train/train/train-from-scratch) for more details.
+
+    To load a model, you can use inference:
+
+    ```python
+    from inference import get_model
+
+    model_id = "<PROJECT_NAME>/<MODEL_VERSION>"
+    model = get_model(model_id=model_id)
+    ```
+
+=== "Ultralytics"
+
+    Similarly to Inference, Ultralytics allows you to run a variety of models.
+
+    ```bash
+    pip install "ultralytics<=8.3.40"
+    ```
+
+    ```python
+    from ultralytics import YOLO
+
+    model = YOLO("yolo11s-seg.pt")
+    ```
+
+## Benchmarking Basics
+
+Evaluating your model requires careful selection of the dataset. Which images should you use?Let's go over the different scenarios.
+
+- **Unrelated Dataset**: If you have a dataset that was not used to train the model, this is the best choice.
+- **Training Set**: This is the set of images used to train the model. This is fine if the model was not trained on this dataset. Otherwise, **never** use it for benchmarking - the results will seem unrealistically good.
+- **Validation Set**: This is the set of images used to validate the model during training. Every Nth training epoch, the model is evaluated on the validation set. Often the training is stopped once the validation loss stops improving. Therefore, even while the images aren't used to train the model, it still indirectly influences the training outcome.
+- **Test Set**: This is the set of images kept aside for model testing. It is exactly the set you should use for benchmarking. If the dataset was split correctly, none of these images would be shown to the model during training.
+
+Therefore, an unrelated dataset or the `test` set is the best choice for benchmarking.
+Several other problems may arise:
+
+- **Extra Classes**: An unrelated dataset may contain additional classes which you may need to [filter out](https://supervision.roboflow.com/how_to/filter_detections/#by-set-of-classes) before computing metrics.
+- **Class Mismatch**: In an unrelated dataset, the class names or IDs may be different to what your model produces, you'll need to remap them, which is [shown in this guide](#running-a-model).
+- **Data Contamination**: The `test` set may not be split correctly, with images from the test set also present in `training` or `validation` set and used during training. In this case, the results will be overly optimistic. This also applies when **very similar** images are used for training and testing - e.g. those taken in the same environment, same lighting conditions, similar angle, etc.
+- **Missing Test Set**: Some datasets do not come with a test set. In this case, you should collect and [label](https://roboflow.com/annotate) your own data. Alternatively, a validation set could be used, but the results could be overly optimistic. Make sure to test in the real world as soon as possible.
+
+## Running a Model
+
+At this stage, you should have:
+
+- A dataset of labeled images to evaluate the model.
+- A model prepared for benchmarking.
+
+With these ready, we can now run the model and obtain predictions.
+We'll use `supervision` to create a dataset iterator, and then run the model on each image.
+
+=== "Inference"
+
+    ```python
+    import supervision as sv
+
+    test_set = 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"
+    )
+
+    image_paths = []
+    predictions_list = []
+    targets_list = []
+
+    for image_path, image, label in test_set:
+        result = model.infer(image)[0]
+        predictions = sv.Detections.from_inference(result)
+
+        image_paths.append(image_path)
+        predictions_list.append(predictions)
+        targets_list.append(label)
+    ```
+
+=== "Ultralytics"
+
+    ```python
+    import supervision as sv
+
+    test_set = 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"
+    )
+
+    image_paths = []
+    predictions_list = []
+    targets_list = []
+
+    for image_path, image, label in test_set:
+        result = model(image)[0]
+        predictions = sv.Detections.from_ultralytics(result)
+
+        image_paths.append(image_path)
+        predictions_list.append(predictions)
+        targets_list.append(label)
+    ```
+
+## Remapping classes
+
+Did you notice an issue in the above logic?
+Since we're using an unrelated dataset, the class names and IDs may be different from what the model was trained on.
+
+We need to remap them to match the dataset classes. Here's how to do it:
+
+```python
+def remap_classes(
+    detections: sv.Detections,
+    class_ids_from_to: dict[int, int],
+    class_names_from_to: dict[str, str]
+) -> None:
+    new_class_ids = [
+        class_ids_from_to.get(class_id, class_id) for class_id in detections.class_id]
+    detections.class_id = np.array(new_class_ids)
+
+    new_class_names = [
+        class_names_from_to.get(name, name) for name in detections["class_name"]]
+    predictions["class_name"] = np.array(new_class_names)
+```
+
+Let's also remove the predictions that are not in the dataset classes.
+
+=== "Inference"
+
+    Dataset class names and IDs can be found in the `data.yaml` file, or by printing `dataset.classes`.
+
+    ```python
+    import supervision as sv
+
+    test_set = 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"
+    )
+
+    image_paths = []
+    predictions_list = []
+    targets_list = []
+
+    for image_path, image, label in test_set:
+        result = model.infer(image)[0]
+        predictions = sv.Detections.from_inference(result)
+
+        remap_classes(
+            detections=predictions,
+            class_ids_from_to={16: 0},
+            class_names_from_to={"dog": "Corgi"}
+        )
+        predictions = predictions[
+            np.isin(predictions["class_name"], test_set.classes)
+        ]
+
+        image_paths.append(image_path)
+        predictions_list.append(predictions)
+        targets_list.append(label)
+    ```
+
+=== "Ultralytics"
+
+    Dataset class names and IDs can be found in the `data.yaml` file, or by printing `dataset.classes`.
+
+    Each model will have a different class mapping, so make sure to check the model's documentation. In this case, the model was trained on the COCO dataset, with a class
+    configuration found [here](https://github.com/ultralytics/ultralytics/blob/main/ultralytics/cfg/datasets/coco8.yaml).
+
+    ```python
+    import supervision as sv
+
+    test_set = 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"
+    )
+
+    image_paths = []
+    predictions_list = []
+    targets_list = []
+
+    for image_path, image, label in test_set:
+        result = model(image)[0]
+        predictions = sv.Detections.from_ultralytics(result)
+
+        remap_classes(
+            detections=predictions,
+            class_ids_from_to={16: 0},
+            class_names_from_to={"dog": "Corgi"}
+        )
+        predictions = predictions[
+            np.isin(predictions["class_name"], test_set.classes)
+        ]
+
+        image_paths.append(image_path)
+        predictions_list.append(predictions)
+        targets_list.append(label)
+    ```
+
+## Visualizing Predictions
+
+The first step in evaluating your model’s performance is to visualize its predictions.
+This gives an intuitive sense of how well your model is detecting objects and where it might be failing.
+
+```python
+import supervision as sv
+
+N = 9
+GRID_SIZE = (3, 3)
+
+target_annotator = sv.PolygonAnnotator(color=sv.Color.from_hex("#8315f9"), thickness=8)
+prediction_annotator = sv.PolygonAnnotator(color=sv.Color.from_hex("#00cfc6"), thickness=6)
+
+
+annotated_images = []
+for image_path, predictions, targets in zip(
+  image_paths[:N], predictions_list[:N], targets_list[:N]
+):
+    annotated_image = cv2.imread(image_path)
+    annotated_image = target_annotator.annotate(scene=annotated_image, detections=targets)
+    annotated_image = prediction_annotator.annotate(scene=annotated_image, detections=prediction)
+    annotated_images.append(annotated_image)
+
+sv.plot_images_grid(images=annotated_images, grid_size=GRID_SIZE)
+```
+
+Here, predictions in purple are targets (ground truth), and predictions in teal are model predictions.
+
+![Basic Model Comparison](https://media.roboflow.com/supervision/image-examples/how-to/benchmark-models/basic-model-comparison-corgi.png)
+
+!!! tip
+
+    Use `sv.BoxAnnotator` for object detection and `sv.OrientedBoxAnnotator` for OBB.
+
+    See [annotator documentation](https://supervision.roboflow.com/latest/detection/annotators/) for even more options.
+
+## Benchmarking Metrics
+
+With multiple models, fine details matter. Visual inspection may not be enough. `supervision` provides a collection of metrics that help obtain precise numerical results of model performance.
+
+### Mean Average Precision (mAP)
+
+We'll start with [MeanAveragePrecision (mAP)](https://supervision.roboflow.com/latest/metrics/mean_average_precision/#supervision.metrics.mean_average_precision.MeanAveragePrecision), which is the most commonly used metric for object detection. It measures the average precision across all classes and IoU thresholds.
+
+For a thorough explanation, check out our [blog](https://blog.roboflow.com/mean-average-precision/) and [Youtube video](https://www.youtube.com/watch?v=oqXDdxF_Wuw).
+
+Here, the most popular value is `mAP 50:95`. It represents the average precision across all classes and IoU thresholds (`0.5` to `0.95`), whereas other values such as `mAP 50` or `mAP 75` only consider a single IoU threshold (`0.5` and `0.75` respectively).
+
+Let's compute the mAP:
+
+```python
+from supervision.metrics import MeanAveragePrecision, MetricTarget
+
+map_metric = MeanAveragePrecision(metric_target=MetricTarget.MASKS)
+map_result = map_metric.update(predictions_list, targets_list).compute()
+```
+
+Try printing the result to see it at a glance:
+
+```python
+print(map_result)
+```
+
+```
+MeanAveragePrecisionResult:
+Metric target: MetricTarget.MASKS
+Class agnostic: False
+mAP @ 50:95: 0.2409
+mAP @ 50:    0.3591
+mAP @ 75:    0.2915
+mAP scores: [0.35909 0.3468 0.34556 ...]
+IoU thresh: [0.5 0.55 0.6 ...]
+AP per class:
+  0: [0.35909 0.3468 0.34556 ...]
+...
+Small objects: ...
+Medium objects: ...
+Large objects: ...
+```
+
+You can also plot the results:
+
+```python
+map_result.plot()
+```
+
+![mAP Plot](https://media.roboflow.com/supervision/image-examples/how-to/benchmark-models/mAP-plot-corgi.png)
+
+The metric also breaks down the results by detected object area. Small, medium and large are simply those with area less than 32², between 32² and 96², and greater than 96² pixels respectively.
+
+### F1 Score
+
+The [F1 Score](https://supervision.roboflow.com/latest/metrics/f1_score/) is another useful metric, especially when dealing with an imbalance between false positives and false negatives. It’s the harmonic mean of **precision** (how many predictions are correct) and **recall** (how many actual instances were detected).
+
+Here's how you can compute the F1 score:
+
+```python
+from supervision.metrics import F1Score, MetricTarget
+
+f1_metric = F1Score(metric_target=MetricTarget.MASKS)
+f1_result = f1_metric.update(predictions_list, targets_list).compute()
+```
+
+As with mAP, you can also print the result:
+
+```python
+print(f1_result)
+```
+
+```
+F1ScoreResult:
+Metric target: MetricTarget.MASKS
+Averaging method: AveragingMethod.WEIGHTED
+F1 @ 50:     0.5341
+F1 @ 75:     0.4636
+F1 @ thresh: [0.53406 0.5278 0.52153 ...]
+IoU thresh: [0.5 0.55 0.6 ...]
+F1 per class:
+  0: [0.53406 0.5278 0.52153 ...]
+...
+Small objects: ...
+Medium objects: ...
+Large objects: ...
+```
+
+Similarly, you can plot the results:
+
+```python
+f1_result.plot()
+```
+
+![F1 Plot](https://media.roboflow.com/supervision/image-examples/how-to/benchmark-models/f1-score-corgi.png)
+
+As with mAP, the metric also breaks down the results by detected object area. Small, medium and large are simply those with area less than 32², between 32² and 96², and greater than 96² pixels respectively.
+
+## Model Leaderboard
+
+Here to compare the basic models? We've got you covered. Check out our [Model Leaderboard](https://leaderboard.roboflow.com/) to see how different models perform and to get a sense of the state-of-the-art results. It's a great place to understand what the leading models can achieve and to compare your own results.
+
+Even better, the repository is open source! You can see how the models were benchmarked, run the evaluation yourself, and even add your own models to the leaderboard. Check it out on [GitHub](https://github.com/roboflow/model-leaderboard)!
+
+![Model Leaderboard Example](https://media.roboflow.com/model-leaderboard/model-leaderboard-example.png)
+
+## Conclusion
+
+In this guide, you've learned how to set up your environment, train or use pre-trained models, visualize predictions, and evaluate model performance with metrics like [mAP](https://supervision.roboflow.com/latest/metrics/mean_average_precision/), [F1 score](https://supervision.roboflow.com/latest/metrics/f1_score/), and got to know our Model Leaderboard.
+
+A condensed version of this guide is also available as a [Colab Notebook](https://colab.research.google.com/drive/1HoOY9pZoVwGiRMmLHtir0qT6Uj45w6Ps?usp=sharing).
+
+For more details, be sure to check out our [documentation](https://supervision.roboflow.com/latest/) and join our community discussions. If you find any issues, please let us know on [GitHub](https://github.com/roboflow/supervision/issues).
+
+Best of luck with your benchmarking!

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

@@ -33,8 +33,9 @@ We write your reusable computer vision tools. Whether you need to load your data
 You can install `supervision` in a
 [**Python>=3.8**](https://www.python.org/) environment.
 
-!!! example "pip install (recommended)"
-    === "pip"
+!!! example "Installation"
+
+    === "pip (recommended)"
         [![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)
         [![license](https://img.shields.io/pypi/l/supervision)](https://github.com/roboflow/supervision/blob/main/LICENSE.md)
@@ -44,6 +45,43 @@ You can install `supervision` in a
         pip install supervision
         ```
 
+    === "poetry"
+        [![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)
+        [![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)
+
+        ```bash
+        poetry add supervision
+        ```
+
+    === "uv"
+        [![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)
+        [![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)
+
+        ```bash
+        uv pip install supervision
+        ```
+
+        For uv projects:
+
+        ```bash
+        uv add supervision
+        ```
+
+    === "rye"
+        [![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)
+        [![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)
+
+        ```bash
+        rye add supervision
+        ```
+
+
 !!! example "conda/mamba install"
     === "conda"
         [![conda-recipe](https://img.shields.io/badge/recipe-supervision-green.svg)](https://anaconda.org/conda-forge/supervision) [![conda-downloads](https://img.shields.io/conda/dn/conda-forge/supervision.svg)](https://anaconda.org/conda-forge/supervision) [![conda-version](https://img.shields.io/conda/vn/conda-forge/supervision.svg)](https://anaconda.org/conda-forge/supervision) [![conda-platforms](https://img.shields.io/conda/pn/conda-forge/supervision.svg)](https://anaconda.org/conda-forge/supervision)
@@ -63,7 +101,7 @@ You can install `supervision` in a
     === "virtualenv"
         ```bash
         # clone repository and navigate to root directory
-        git clone https://github.com/roboflow/supervision.git
+        git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
         cd supervision
 
         # setup python environment and activate it
@@ -75,18 +113,19 @@ You can install `supervision` in a
         pip install -e "."
         ```
 
-    === "poetry"
+    === "uv"
         ```bash
         # clone repository and navigate to root directory
-        git clone https://github.com/roboflow/supervision.git
+        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

docs/javascripts/segment.js

@@ -0,0 +1,4 @@
+!function(){var i="analytics",analytics=window[i]=window[i]||[];if(!analytics.initialize)if(analytics.invoked)window.console&&console.error&&console.error("Segment snippet included twice.");else{analytics.invoked=!0;analytics.methods=["trackSubmit","trackClick","trackLink","trackForm","pageview","identify","reset","group","track","ready","alias","debug","page","screen","once","off","on","addSourceMiddleware","addIntegrationMiddleware","setAnonymousId","addDestinationMiddleware","register"];analytics.factory=function(e){return function(){if(window[i].initialized)return window[i][e].apply(window[i],arguments);var n=Array.prototype.slice.call(arguments);if(["track","screen","alias","group","page","identify"].indexOf(e)>-1){var c=document.querySelector("link[rel='canonical']");n.push({__t:"bpc",c:c&&c.getAttribute("href")||void 0,p:location.pathname,u:location.href,s:location.search,t:document.title,r:document.referrer})}n.unshift(e);analytics.push(n);return analytics}};for(var n=0;n<analytics.methods.length;n++){var key=analytics.methods[n];analytics[key]=analytics.factory(key)}analytics.load=function(key,n){var t=document.createElement("script");t.type="text/javascript";t.async=!0;t.setAttribute("data-global-segment-analytics-key",i);t.src="https://cdn.segment.com/analytics.js/v1/" + key + "/analytics.min.js";var r=document.getElementsByTagName("script")[0];r.parentNode.insertBefore(t,r);analytics._loadOptions=n};analytics._writeKey="rMvrPeZBJYyOPJSGCNhMlnTJb8VhFiWU";;analytics.SNIPPET_VERSION="5.2.0";
+  analytics.load("eohFog7VZiAhGJGEr5Sh7BM1mFKmUvDC");
+  document$.subscribe(analytics.page);
+  }}();

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

@@ -130,7 +130,7 @@
             },
             "outputs": [],
             "source": [
-                "!pip install -q ultralytics supervision==0.18.0"
+                "!pip install -q \"ultralytics<=8.3.40\" \"supervision==0.18.0\""
             ]
         },
         {

docs/notebooks/quickstart.ipynb

@@ -300,7 +300,7 @@
             },
             "outputs": [],
             "source": [
-                "!pip install -q ultralytics"
+                "!pip install -q \"ultralytics<=8.3.40\""
             ]
         },
         {

examples/count_people_in_zone/README.md

@@ -17,7 +17,7 @@ https://github.com/roboflow/supervision/assets/26109316/f84db7b5-79e2-4142-a1da-
 - clone repository and navigate to example directory
 
     ```bash
-    git clone https://github.com/roboflow/supervision.git
+    git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
     cd supervision/examples/count_people_in_zone
     ```
 

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/requirements.txt

@@ -2,4 +2,4 @@ gdown
 inference==0.9.17
 supervision>=0.20.0
 tqdm
-ultralytics
+ultralytics<=8.3.40

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/heatmap_and_track/README.md

@@ -11,7 +11,7 @@ supervision package for multiple tasks such as drawing heatmap annotations, trac
 - clone repository and navigate to example directory
 
     ```bash
-    git clone https://github.com/roboflow/supervision.git
+    git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
     cd supervision/examples/heatmap_and_track
     ```
 

examples/heatmap_and_track/requirements.txt

@@ -1,2 +1,2 @@
 supervision[assets]==0.19.0
-ultralytics
+ultralytics<=8.3.40

examples/speed_estimation/README.md

@@ -22,7 +22,7 @@ https://github.com/roboflow/supervision/assets/26109316/d50118c1-2ae4-458d-915a-
 - clone repository and navigate to example directory
 
     ```bash
-    git clone https://github.com/roboflow/supervision.git
+    git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
     cd supervision/examples/speed_estimation
     ```
 

examples/time_in_zone/README.md

@@ -16,7 +16,7 @@ https://github.com/roboflow/supervision/assets/26109316/d051cc8a-dd15-41d4-aa36-
 - clone repository and navigate to example directory
 
     ```bash
-    git clone https://github.com/roboflow/supervision.git
+    git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
     cd supervision/examples/time_in_zone
     ```
 

examples/time_in_zone/requirements.txt

@@ -1,5 +1,5 @@
 opencv-python
 supervision>=0.20.0
-ultralytics
+ultralytics<=8.3.40
 inference==0.9.17
 pytube

examples/tracking/README.md

@@ -10,7 +10,7 @@ detection and Supervision for tracking and annotation.
 - clone repository and navigate to example directory
 
     ```bash
-    git clone https://github.com/roboflow/supervision.git
+    git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
     cd supervision/examples/tracking
     ```
 

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/requirements.txt

@@ -1,4 +1,4 @@
 inference==0.9.17
 supervision==0.19.0
 tqdm
-ultralytics
+ultralytics<=8.3.40

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)

examples/traffic_analysis/README.md

@@ -13,7 +13,7 @@ https://github.com/roboflow/supervision/assets/26109316/c9436828-9fbf-4c25-ae8c-
 - clone repository and navigate to example directory
 
     ```bash
-    git clone https://github.com/roboflow/supervision.git
+    git clone --depth 1 -b develop https://github.com/roboflow/supervision.git
     cd supervision/examples/traffic_analysis
     ```
 

examples/traffic_analysis/requirements.txt

@@ -2,4 +2,4 @@ gdown
 inference==0.9.17
 supervision>=0.20.0
 tqdm
-ultralytics
+ultralytics<=8.3.40

mkdocs.yml

@@ -42,6 +42,7 @@ nav:
       - 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
+      - Benchmark a Model: how_to/benchmark_a_model.md
 
   - Reference - Code API:
       - Detection and Segmentation:
@@ -148,9 +149,12 @@ plugins:
             group_by_category: true
             docstring_style: google
             show_symbol_type_heading: true
+            show_root_heading: True
             show_symbol_type_toc: true
             show_category_heading: true
-          domains: [std, py]
+          inventories:
+          - url: https://docs.python-requests.org/en/master/objects.inv
+            domains: [std, py]
   - git-committers:
       repository: roboflow/supervision
       branch: develop
@@ -175,11 +179,13 @@ markdown_extensions:
       check_paths: true
   - pymdownx.highlight:
       anchor_linenums: true
+      line_spans: __span
+      pygments_lang_class: true
 
 extra_javascript:
-  - "https://widget.kapa.ai/kapa-widget.bundle.js"
   - "javascripts/init_kapa_widget.js"
   - "javascripts/cookbooks-card.js"
+  - "javascripts/segment.js"
   - "https://cdnjs.cloudflare.com/ajax/libs/dompurify/3.0.8/purify.min.js"
 
 # Messages shown during document build

pyproject.toml

@@ -1,18 +1,16 @@
-[tool.poetry]
+[project]
 name = "supervision"
-version = "0.25.0"
 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 = "0.26.0rc7"
 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',
@@ -40,95 +37,59 @@ classifiers = [
     'Operating System :: POSIX :: Linux',
     'Operating System :: MacOS',
 ]
-
-[tool.poetry.dependencies]
-python = "^3.8"
-numpy = [
-    { version = ">=1.21.2,<1.23.3", python = "<=3.10" },
-    { version = ">=1.23.3", python = ">3.10" },
-    { version = ">=2.1.0", python = ">=3.13" },
-]
-scipy = [
-    { version = "1.10.0", python = "<3.9" },
-    { version = "^1.10.0", python = ">=3.9" },
-    { version = ">=1.14.1", 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"
 ]
 
-# 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" },
-    { version = ">=1.3.0", 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 = ">=3.6.0"
-pyyaml = ">=5.3"
-defusedxml = "^0.7.1"
-pillow = ">=9.4"
-requests = { version = ">=2.26.0,<=2.32.3", optional = true }
-tqdm = { version = ">=4.62.3,<=4.67.0", 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 }
-opencv-python = ">=4.5.5.64"
-
-[tool.poetry.extras]
-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,<0.46"
-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" },
+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"
+]
+docs = [
+    "mkdocs-material[imaging]>=9.5.5",
+    "mkdocstrings>=0.25.2,<0.30.0",
+    "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"
+]
+build = [
+    "twine>=5.1.1,<7.0.0",
+    "wheel>=0.40,<0.46",
+    "build>=0.10,<1.3"
 ]
-
-[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"
-
 
 [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"
 
@@ -252,7 +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"
@@ -261,8 +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

@@ -9,11 +9,11 @@ except importlib_metadata.PackageNotFoundError:
 from supervision.annotators.core import (
     BackgroundOverlayAnnotator,
     BlurAnnotator,
-    BoundingBoxAnnotator,
     BoxAnnotator,
     BoxCornerAnnotator,
     CircleAnnotator,
     ColorAnnotator,
+    ComparisonAnnotator,
     CropAnnotator,
     DotAnnotator,
     EllipseAnnotator,
@@ -45,7 +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,
@@ -76,7 +75,11 @@ from supervision.detection.utils import (
     scale_boxes,
     xcycwh_to_xyxy,
     xywh_to_xyxy,
+    xyxy_to_polygons,
+    xyxy_to_xcycarh,
+    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,
@@ -118,3 +121,111 @@ from supervision.utils.video import (
     get_video_frames_generator,
     process_video,
 )
+
+__all__ = [
+    "LMM",
+    "BackgroundOverlayAnnotator",
+    "BaseDataset",
+    "BlurAnnotator",
+    "BoxAnnotator",
+    "BoxCornerAnnotator",
+    "ByteTrack",
+    "CSVSink",
+    "CircleAnnotator",
+    "ClassificationDataset",
+    "Classifications",
+    "Color",
+    "ColorAnnotator",
+    "ColorLookup",
+    "ColorPalette",
+    "ComparisonAnnotator",
+    "ConfusionMatrix",
+    "CropAnnotator",
+    "DetectionDataset",
+    "Detections",
+    "DetectionsSmoother",
+    "DotAnnotator",
+    "EdgeAnnotator",
+    "EllipseAnnotator",
+    "FPSMonitor",
+    "HaloAnnotator",
+    "HeatMapAnnotator",
+    "IconAnnotator",
+    "ImageSink",
+    "InferenceSlicer",
+    "JSONSink",
+    "KeyPoints",
+    "LabelAnnotator",
+    "LineZone",
+    "LineZoneAnnotator",
+    "LineZoneAnnotatorMulticlass",
+    "MaskAnnotator",
+    "MeanAveragePrecision",
+    "OrientedBoxAnnotator",
+    "OverlapFilter",
+    "PercentageBarAnnotator",
+    "PixelateAnnotator",
+    "Point",
+    "PolygonAnnotator",
+    "PolygonZone",
+    "PolygonZoneAnnotator",
+    "Position",
+    "Rect",
+    "RichLabelAnnotator",
+    "RoundBoxAnnotator",
+    "TraceAnnotator",
+    "TriangleAnnotator",
+    "VertexAnnotator",
+    "VertexLabelAnnotator",
+    "VideoInfo",
+    "VideoSink",
+    "box_iou_batch",
+    "box_non_max_merge",
+    "box_non_max_suppression",
+    "calculate_masks_centroids",
+    "calculate_optimal_line_thickness",
+    "calculate_optimal_text_scale",
+    "clip_boxes",
+    "contains_holes",
+    "contains_multiple_segments",
+    "create_tiles",
+    "crop_image",
+    "cv2_to_pillow",
+    "draw_filled_polygon",
+    "draw_filled_rectangle",
+    "draw_image",
+    "draw_line",
+    "draw_polygon",
+    "draw_rectangle",
+    "draw_text",
+    "filter_polygons_by_area",
+    "get_polygon_center",
+    "get_video_frames_generator",
+    "letterbox_image",
+    "list_files_with_extensions",
+    "mask_iou_batch",
+    "mask_non_max_suppression",
+    "mask_to_polygons",
+    "mask_to_rle",
+    "mask_to_xyxy",
+    "move_boxes",
+    "move_masks",
+    "oriented_box_iou_batch",
+    "overlay_image",
+    "pad_boxes",
+    "pillow_to_cv2",
+    "plot_image",
+    "plot_images_grid",
+    "polygon_to_mask",
+    "polygon_to_xyxy",
+    "process_video",
+    "resize_image",
+    "rle_to_mask",
+    "scale_boxes",
+    "scale_image",
+    "xcycwh_to_xyxy",
+    "xywh_to_xyxy",
+    "xyxy_to_polygons",
+    "xyxy_to_xyah",
+    "xyxy_to_xywh",
+]

supervision/annotators/core.py

@@ -9,6 +9,7 @@ from PIL import Image, ImageDraw, ImageFont
 
 from supervision.annotators.base import BaseAnnotator, ImageType
 from supervision.annotators.utils import (
+    PENDING_TRACK_ID,
     ColorLookup,
     Trace,
     resolve_color,
@@ -16,10 +17,16 @@ from supervision.annotators.utils import (
 )
 from supervision.config import CLASS_NAME_DATA_FIELD, ORIENTED_BOX_COORDINATES
 from supervision.detection.core import Detections
-from supervision.detection.utils import clip_boxes, mask_to_polygons, spread_out_boxes
+from supervision.detection.utils import (
+    clip_boxes,
+    mask_to_polygons,
+    polygon_to_mask,
+    spread_out_boxes,
+    xyxy_to_polygons,
+)
 from supervision.draw.color import Color, ColorPalette
-from supervision.draw.utils import draw_polygon
-from supervision.geometry.core import Position
+from supervision.draw.utils import draw_polygon, draw_rounded_rectangle, draw_text
+from supervision.geometry.core import Point, Position, Rect
 from supervision.utils.conversion import (
     ensure_cv2_image_for_annotation,
     ensure_pil_image_for_annotation,
@@ -30,7 +37,6 @@ from supervision.utils.image import (
     overlay_image,
     scale_image,
 )
-from supervision.utils.internal import deprecated
 
 CV2_FONT = cv2.FONT_HERSHEY_SIMPLEX
 
@@ -118,92 +124,6 @@ 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.
@@ -1211,7 +1131,7 @@ class LabelAnnotator(BaseAnnotator):
         detections: Detections, custom_labels: Optional[List[str]]
     ) -> List[str]:
         if custom_labels is not None:
-            return custom_labels
+            return [str(label) for label in custom_labels]
 
         labels = []
         for idx in range(len(detections)):
@@ -1820,6 +1740,7 @@ class TraceAnnotator(BaseAnnotator):
                 "The `tracker_id` field is missing in the provided detections."
                 " See more: https://supervision.roboflow.com/latest/how_to/track_objects"
             )
+        detections = detections[detections.tracker_id != PENDING_TRACK_ID]
 
         self.trace.put(detections)
         for detection_idx in range(len(detections)):
@@ -1938,11 +1859,9 @@ class HeatMapAnnotator(BaseAnnotator):
         temp = temp.astype(np.uint8)
         if self.kernel_size is not None:
             temp = cv2.blur(temp, (self.kernel_size, self.kernel_size))
-        hsv = np.zeros(scene.shape)
+        hsv = np.full(scene.shape, 255, dtype=np.uint8)
         hsv[..., 0] = temp
-        hsv[..., 1] = 255
-        hsv[..., 2] = 255
-        temp = cv2.cvtColor(hsv.astype(np.uint8), cv2.COLOR_HSV2BGR)
+        temp = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)
         mask = cv2.cvtColor(self.heat_mask.astype(np.uint8), cv2.COLOR_GRAY2BGR) > 0
         scene[mask] = cv2.addWeighted(temp, self.opacity, scene, 1 - self.opacity, 0)[
             mask
@@ -2527,6 +2446,9 @@ class CropAnnotator(BaseAnnotator):
                 detections=detections
             )
             ```
+
+        ![crop-annotator-example](https://media.roboflow.com/
+        supervision-annotator-examples/crop-annotator-example.png)
         """
         assert isinstance(scene, np.ndarray)
         crops = [
@@ -2683,3 +2605,261 @@ class BackgroundOverlayAnnotator(BaseAnnotator):
 
         np.copyto(scene, colored_mask)
         return scene
+
+
+class ComparisonAnnotator:
+    """
+    Highlights the differences between two sets of detections.
+    Useful for comparing results from two different models, or the difference
+    between a ground truth and a prediction.
+
+    If present, uses the oriented bounding box data.
+    Otherwise, if present, uses a mask.
+    Otherwise, uses the bounding box data.
+    """
+
+    def __init__(
+        self,
+        color_1: Color = Color.RED,
+        color_2: Color = Color.GREEN,
+        color_overlap: Color = Color.BLUE,
+        *,
+        opacity: float = 0.75,
+        label_1: str = "",
+        label_2: str = "",
+        label_overlap: str = "",
+        label_scale: float = 1.0,
+    ):
+        """
+        Args:
+            color_1 (Color): Color of areas only present in the first set of
+                detections.
+            color_2 (Color): Color of areas only present in the second set of
+                detections.
+            color_overlap (Color): Color of areas present in both sets of detections.
+            opacity (float): Annotator opacity, from `0` to `1`.
+            label_1 (str): Label for the first set of detections.
+            label_2 (str): Label for the second set of detections.
+            label_overlap (str): Label for areas present in both sets of detections.
+            label_scale (float): Controls how large the labels are.
+        """
+
+        self.color_1 = color_1
+        self.color_2 = color_2
+        self.color_overlap = color_overlap
+
+        self.opacity = opacity
+        self.label_1 = label_1
+        self.label_2 = label_2
+        self.label_overlap = label_overlap
+        self.label_scale = label_scale
+        self.text_thickness = int(self.label_scale + 1.2)
+
+    @ensure_cv2_image_for_annotation
+    def annotate(
+        self, scene: ImageType, detections_1: Detections, detections_2: Detections
+    ) -> ImageType:
+        """
+        Highlights the differences between two sets of detections.
+
+        Args:
+            scene (ImageType): The image where detections will be drawn.
+                `ImageType` is a flexible type, accepting either `numpy.ndarray`
+                or `PIL.Image.Image`.
+            detections_1 (Detections): The first set of detections or predictions.
+            detections_2 (Detections): The second set of detections to compare or
+                ground truth.
+
+        Returns:
+            The annotated image.
+
+        Example:
+            ```python
+            import supervision as sv
+
+            image = ...
+            detections_1 = sv.Detections(...)
+            detections_2 = sv.Detections(...)
+
+            comparison_annotator = sv.ComparisonAnnotator()
+            annotated_frame = comparison_annotator.annotate(
+                scene=image.copy(),
+                detections_1=detections_1,
+                detections_2=detections_2
+            )
+            ```
+
+        ![comparison-annotator-example](https://media.roboflow.com/
+        supervision-annotator-examples/comparison-annotator-example.png)
+        """
+        assert isinstance(scene, np.ndarray)
+        if detections_1.is_empty() and detections_2.is_empty():
+            return scene
+
+        use_obb = self._use_obb(detections_1, detections_2)
+        use_mask = self._use_mask(detections_1, detections_2)
+
+        if use_obb:
+            mask_1 = self._mask_from_obb(scene, detections_1)
+            mask_2 = self._mask_from_obb(scene, detections_2)
+
+        elif use_mask:
+            mask_1 = self._mask_from_mask(scene, detections_1)
+            mask_2 = self._mask_from_mask(scene, detections_2)
+
+        else:
+            mask_1 = self._mask_from_xyxy(scene, detections_1)
+            mask_2 = self._mask_from_xyxy(scene, detections_2)
+
+        mask_overlap = mask_1 & mask_2
+        mask_1 = mask_1 & ~mask_overlap
+        mask_2 = mask_2 & ~mask_overlap
+
+        color_layer = np.zeros_like(scene, dtype=np.uint8)
+        color_layer[mask_overlap] = self.color_overlap.as_bgr()
+        color_layer[mask_1] = self.color_1.as_bgr()
+        color_layer[mask_2] = self.color_2.as_bgr()
+
+        scene[mask_overlap] = (1 - self.opacity) * scene[
+            mask_overlap
+        ] + self.opacity * color_layer[mask_overlap]
+        scene[mask_1] = (1 - self.opacity) * scene[mask_1] + self.opacity * color_layer[
+            mask_1
+        ]
+        scene[mask_2] = (1 - self.opacity) * scene[mask_2] + self.opacity * color_layer[
+            mask_2
+        ]
+
+        self._draw_labels(scene)
+
+        return scene
+
+    @staticmethod
+    def _use_obb(detections_1: Detections, detections_2: Detections) -> bool:
+        assert not detections_1.is_empty() or not detections_2.is_empty()
+        is_obb_1 = ORIENTED_BOX_COORDINATES in detections_1.data
+        is_obb_2 = ORIENTED_BOX_COORDINATES in detections_2.data
+        return (
+            (is_obb_1 and is_obb_2)
+            or (is_obb_1 and detections_2.is_empty())
+            or (detections_1.is_empty() and is_obb_2)
+        )
+
+    @staticmethod
+    def _use_mask(detections_1: Detections, detections_2: Detections) -> bool:
+        assert not detections_1.is_empty() or not detections_2.is_empty()
+        is_mask_1 = detections_1.mask is not None
+        is_mask_2 = detections_2.mask is not None
+        return (
+            (is_mask_1 and is_mask_2)
+            or (is_mask_1 and detections_2.is_empty())
+            or (detections_1.is_empty() and is_mask_2)
+        )
+
+    @staticmethod
+    def _mask_from_xyxy(scene: np.ndarray, detections: Detections) -> np.ndarray:
+        mask = np.zeros(scene.shape[:2], dtype=np.bool_)
+        if detections.is_empty():
+            return mask
+
+        resolution_wh = scene.shape[1], scene.shape[0]
+        polygons = xyxy_to_polygons(detections.xyxy)
+
+        for polygon in polygons:
+            polygon_mask = polygon_to_mask(polygon, resolution_wh=resolution_wh)
+            mask |= polygon_mask.astype(np.bool_)
+        return mask
+
+    @staticmethod
+    def _mask_from_obb(scene: np.ndarray, detections: Detections) -> np.ndarray:
+        mask = np.zeros(scene.shape[:2], dtype=np.bool_)
+        if detections.is_empty():
+            return mask
+
+        resolution_wh = scene.shape[1], scene.shape[0]
+
+        for polygon in detections.data[ORIENTED_BOX_COORDINATES]:
+            polygon_mask = polygon_to_mask(polygon, resolution_wh=resolution_wh)
+            mask |= polygon_mask.astype(np.bool_)
+        return mask
+
+    @staticmethod
+    def _mask_from_mask(scene: np.ndarray, detections: Detections) -> np.ndarray:
+        mask = np.zeros(scene.shape[:2], dtype=np.bool_)
+        if detections.is_empty():
+            return mask
+        assert detections.mask is not None
+
+        for detections_mask in detections.mask:
+            mask |= detections_mask.astype(np.bool_)
+        return mask
+
+    def _draw_labels(self, scene: np.ndarray) -> None:
+        """
+        Draw the labels, explaining what each color represents, with automatically
+        computed positions.
+
+        Args:
+            scene (np.ndarray): The image where the labels will be drawn.
+        """
+        margin = int(50 * self.label_scale)
+        gap = int(40 * self.label_scale)
+        y0 = int(50 * self.label_scale)
+        height = int(50 * self.label_scale)
+
+        marker_size = int(20 * self.label_scale)
+        padding = int(10 * self.label_scale)
+        text_box_corner_radius = int(10 * self.label_scale)
+        marker_corner_radius = int(4 * self.label_scale)
+        text_scale = self.label_scale
+
+        label_color_pairs = [
+            (self.label_1, self.color_1),
+            (self.label_2, self.color_2),
+            (self.label_overlap, self.color_overlap),
+        ]
+
+        x0 = margin
+        for text, color in label_color_pairs:
+            if not text:
+                continue
+
+            (text_w, _) = cv2.getTextSize(
+                text=text,
+                fontFace=CV2_FONT,
+                fontScale=self.label_scale,
+                thickness=self.text_thickness,
+            )[0]
+
+            width = text_w + marker_size + padding * 4
+            center_x = x0 + width // 2
+            center_y = y0 + height // 2
+
+            draw_rounded_rectangle(
+                scene=scene,
+                rect=Rect(x=x0, y=y0, width=width, height=height),
+                color=Color.WHITE,
+                border_radius=text_box_corner_radius,
+            )
+
+            draw_rounded_rectangle(
+                scene=scene,
+                rect=Rect(
+                    x=x0 + padding,
+                    y=center_y - marker_size / 2,
+                    width=marker_size,
+                    height=marker_size,
+                ),
+                color=color,
+                border_radius=marker_corner_radius,
+            )
+
+            draw_text(
+                scene,
+                text,
+                text_anchor=Point(x=center_x + marker_size, y=center_y),
+                text_scale=text_scale,
+                text_thickness=self.text_thickness,
+            )
+
+            x0 += width + gap

supervision/annotators/utils.py

@@ -7,6 +7,9 @@ from supervision.detection.core import Detections
 from supervision.draw.color import Color, ColorPalette
 from supervision.geometry.core import Position
 
+PENDING_TRACK_COLOR = Color.GREY
+PENDING_TRACK_ID = -1
+
 
 class ColorLookup(Enum):
     """
@@ -136,6 +139,8 @@ def resolve_color(
         detection_idx=detection_idx,
         color_lookup=color_lookup,
     )
+    if color_lookup == ColorLookup.TRACK and idx == PENDING_TRACK_ID:
+        return PENDING_TRACK_COLOR
     return get_color_by_index(color=color, idx=idx)
 
 

supervision/assets/__init__.py

@@ -1,2 +1,4 @@
 from supervision.assets.downloader import download_assets
 from supervision.assets.list import VideoAssets
+
+__all__ = ["VideoAssets", "download_assets"]

supervision/assets/downloader.py

@@ -4,19 +4,10 @@ from pathlib import Path
 from shutil import copyfileobj
 from typing import Union
 
-from supervision.assets.list import VIDEO_ASSETS, VideoAssets
+from requests import get
+from tqdm.auto import tqdm
 
-try:
-    from requests import get
-    from tqdm.auto import tqdm
-except ImportError:
-    raise ValueError(
-        "\n"
-        "Please install requests and tqdm to download assets \n"
-        "or install supervision with assets \n"
-        "pip install supervision[assets] \n"
-        "\n"
-    )
+from supervision.assets.list import VIDEO_ASSETS, VideoAssets
 
 
 def is_md5_hash_matching(filename: str, original_md5_hash: str) -> bool:

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

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
-from contextlib import suppress
 from dataclasses import dataclass, field
+from enum import Enum
 from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
 
 import numpy as np
@@ -10,12 +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,
@@ -39,8 +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
 
 
@@ -117,7 +119,7 @@ class Detections:
         xyxy (np.ndarray): An array of shape `(n, 4)` containing
             the bounding boxes coordinates in format `[x1, y1, x2, y2]`
         mask: (Optional[np.ndarray]): An array of shape
-            `(n, H, W)` containing the segmentation masks.
+            `(n, H, W)` containing the segmentation masks (`bool` data type).
         confidence (Optional[np.ndarray]): An array of shape
             `(n,)` containing the confidence scores of the detections.
         class_id (Optional[np.ndarray]): An array of shape
@@ -606,8 +608,10 @@ class Detections:
             detections = sv.Detections.from_inference(result)
             ```
         """
-        with suppress(AttributeError):
+        if hasattr(roboflow_result, "dict"):
             roboflow_result = roboflow_result.dict(exclude_none=True, by_alias=True)
+        elif hasattr(roboflow_result, "json"):
+            roboflow_result = roboflow_result.json()
         xyxy, confidence, class_id, masks, trackers, data = process_roboflow_result(
             roboflow_result=roboflow_result
         )
@@ -711,7 +715,7 @@ class Detections:
         """
         if "error" in azure_result:
             raise ValueError(
-                f'Azure API returned an error {azure_result["error"]["message"]}'
+                f"Azure API returned an error {azure_result['error']['message']}"
             )
 
         xyxy, confidences, class_ids = [], [], []
@@ -798,6 +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:
@@ -836,16 +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()
@@ -858,8 +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:
         """
@@ -1026,16 +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],
@@ -1201,6 +1242,8 @@ class Detections:
         """
         if isinstance(index, str):
             return self.data.get(index)
+        if self.is_empty():
+            return self
         if isinstance(index, int):
             index = [index]
         return Detections(
@@ -1303,9 +1346,9 @@ class Detections:
         if len(self) == 0:
             return self
 
-        assert (
-            self.confidence is not None
-        ), "Detections confidence must be given for NMS to be executed."
+        assert self.confidence is not None, (
+            "Detections confidence must be given for NMS to be executed."
+        )
 
         if class_agnostic:
             predictions = np.hstack((self.xyxy, self.confidence.reshape(-1, 1)))
@@ -1359,9 +1402,9 @@ class Detections:
         if len(self) == 0:
             return self
 
-        assert (
-            self.confidence is not None
-        ), "Detections confidence must be given for NMM to be executed."
+        assert self.confidence is not None, (
+            "Detections confidence must be given for NMM to be executed."
+        )
 
         if class_agnostic:
             predictions = np.hstack((self.xyxy, self.confidence.reshape(-1, 1)))

supervision/detection/lmm.py

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

supervision/detection/overlap_filter.py

@@ -226,8 +226,7 @@ def box_non_max_merge(
     for merge_group in merge_groups:
         if len(merge_group) == 0:
             raise ValueError(
-                f"Empty group detected when non-max-merging "
-                f"detections: {merge_groups}"
+                f"Empty group detected when non-max-merging detections: {merge_groups}"
             )
     return merge_groups
 

supervision/detection/utils.py

@@ -11,6 +11,25 @@ from supervision.geometry.core import Vector
 MIN_POLYGON_POINT_COUNT = 3
 
 
+def xyxy_to_polygons(box: np.ndarray) -> np.ndarray:
+    """
+    Convert an array of boxes to an array of polygons.
+    Retains the input datatype.
+
+    Args:
+        box (np.ndarray): An array of boxes (N, 4), where each box is represented as a
+            list of four coordinates in the format `(x_min, y_min, x_max, y_max)`.
+
+    Returns:
+        np.ndarray: An array of polygons (N, 4, 2), where each polygon is
+            represented as a list of four coordinates in the format `(x, y)`.
+    """
+    polygon = np.zeros((box.shape[0], 4, 2), dtype=box.dtype)
+    polygon[:, :, 0] = box[:, [0, 2, 2, 0]]
+    polygon[:, :, 1] = box[:, [1, 1, 3, 3]]
+    return polygon
+
+
 def polygon_to_mask(polygon: np.ndarray, resolution_wh: Tuple[int, int]) -> np.ndarray:
     """Generate a mask from a polygon.
 
@@ -302,6 +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)`
@@ -341,6 +397,56 @@ def xcycwh_to_xyxy(xcycwh: np.ndarray) -> np.ndarray:
     return xyxy
 
 
+def xyxy_to_xcycarh(xyxy: np.ndarray) -> np.ndarray:
+    """
+    Converts bounding box coordinates from `(x_min, y_min, x_max, y_max)`
+    into measurement space to format `(center x, center y, aspect ratio, height)`,
+    where the aspect ratio is `width / height`.
+
+    Args:
+        xyxy (np.ndarray): Bounding box in format `(x1, y1, x2, y2)`.
+            Expected shape is `(N, 4)`.
+    Returns:
+        np.ndarray: Bounding box in format
+            `(center x, center y, aspect ratio, height)`. Shape `(N, 4)`.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xyxy = np.array([
+            [10, 20, 40, 60],
+            [15, 25, 50, 70]
+        ])
+
+        sv.xyxy_to_xcycarh(xyxy=xyxy)
+        # array([
+        #     [25.  , 40.  ,  0.75, 40.  ],
+        #     [32.5 , 47.5 ,  0.77777778, 45.  ]
+        # ])
+        ```
+
+    """
+    if xyxy.size == 0:
+        return np.empty((0, 4), dtype=float)
+
+    x1, y1, x2, y2 = xyxy.T
+    width = x2 - x1
+    height = y2 - y1
+    center_x = x1 + width / 2
+    center_y = y1 + height / 2
+
+    aspect_ratio = np.divide(
+        width,
+        height,
+        out=np.zeros_like(width, dtype=float),
+        where=height != 0,
+    )
+    result = np.column_stack((center_x, center_y, aspect_ratio, height))
+    return result.astype(float)
+
+
 def mask_to_xyxy(masks: np.ndarray) -> np.ndarray:
     """
     Converts a 3D `np.array` of 2D bool masks into a 2D `np.array` of bounding boxes.
@@ -701,25 +807,71 @@ def move_masks(
         masks (npt.NDArray[np.bool_]): A 3D array of binary masks corresponding to the
             predictions. Shape: `(N, H, W)`, where N is the number of predictions, and
             H, W are the dimensions of each mask.
-        offset (npt.NDArray[np.int32]): An array of shape `(2,)` containing non-negative
-            int values `[dx, dy]`.
+        offset (npt.NDArray[np.int32]): An array of shape `(2,)` containing int values
+            `[dx, dy]`. Supports both positive and negative values for bidirectional
+            movement.
         resolution_wh (Tuple[int, int]): The width and height of the desired mask
             resolution.
 
     Returns:
         (npt.NDArray[np.bool_]) repositioned masks, optionally padded to the specified
             shape.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([[[False, False, False, False],
+                         [False, True,  True,  False],
+                         [False, True,  True,  False],
+                         [False, False, False, False]]], dtype=bool)
+
+        offset = np.array([1, 1])
+        sv.move_masks(mask, offset, resolution_wh=(4, 4))
+        # array([[[False, False, False, False],
+        #         [False, False, False, False],
+        #         [False, False,  True,  True],
+        #         [False, False,  True,  True]]], dtype=bool)
+
+        offset = np.array([-2, 2])
+        sv.move_masks(mask, offset, resolution_wh=(4, 4))
+        # array([[[False, False, False, False],
+        #         [False, False, False, False],
+        #         [False, False, False, False],
+        #         [True,  False, False, False]]], dtype=bool)
+        ```
     """
-
-    if offset[0] < 0 or offset[1] < 0:
-        raise ValueError(f"Offset values must be non-negative integers. Got: {offset}")
-
     mask_array = np.full((masks.shape[0], resolution_wh[1], resolution_wh[0]), False)
-    mask_array[
-        :,
-        offset[1] : masks.shape[1] + offset[1],
-        offset[0] : masks.shape[2] + offset[0],
-    ] = masks
+
+    if offset[0] < 0:
+        source_x_start = -offset[0]
+        source_x_end = min(masks.shape[2], resolution_wh[0] - offset[0])
+        destination_x_start = 0
+        destination_x_end = min(resolution_wh[0], masks.shape[2] + offset[0])
+    else:
+        source_x_start = 0
+        source_x_end = min(masks.shape[2], resolution_wh[0] - offset[0])
+        destination_x_start = offset[0]
+        destination_x_end = offset[0] + source_x_end - source_x_start
+
+    if offset[1] < 0:
+        source_y_start = -offset[1]
+        source_y_end = min(masks.shape[1], resolution_wh[1] - offset[1])
+        destination_y_start = 0
+        destination_y_end = min(resolution_wh[1], masks.shape[1] + offset[1])
+    else:
+        source_y_start = 0
+        source_y_end = min(masks.shape[1], resolution_wh[1] - offset[1])
+        destination_y_start = offset[1]
+        destination_y_end = offset[1] + source_y_end - source_y_start
+
+    if source_x_end > source_x_start and source_y_end > source_y_start:
+        mask_array[
+            :,
+            destination_y_start:destination_y_end,
+            destination_x_start:destination_x_end,
+        ] = masks[:, source_y_start:source_y_end, source_x_start:source_x_end]
 
     return mask_array
 

supervision/detection/vlm.py

@@ -0,0 +1,317 @@
+import json
+import re
+from enum import Enum
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import numpy as np
+
+from supervision.detection.utils import polygon_to_mask, polygon_to_xyxy
+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"
+
+
+class VLM(Enum):
+    PALIGEMMA = "paligemma"
+    FLORENCE_2 = "florence_2"
+    QWEN_2_5_VL = "qwen_2_5_vl"
+
+
+RESULT_TYPES: Dict[VLM, type] = {
+    VLM.PALIGEMMA: str,
+    VLM.FLORENCE_2: dict,
+    VLM.QWEN_2_5_VL: str,
+}
+
+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 = [
+    "<OD>",
+    "<CAPTION_TO_PHRASE_GROUNDING>",
+    "<DENSE_REGION_CAPTION>",
+    "<REGION_PROPOSAL>",
+    "<OCR_WITH_REGION>",
+    "<REFERRING_EXPRESSION_SEGMENTATION>",
+    "<REGION_TO_SEGMENTATION>",
+    "<OPEN_VOCABULARY_DETECTION>",
+    "<REGION_TO_CATEGORY>",
+    "<REGION_TO_DESCRIPTION>",
+]
+
+
+def validate_vlm_parameters(
+    vlm: Union[VLM, str], result: Any, kwargs: Dict[str, Any]
+) -> VLM:
+    if isinstance(vlm, str):
+        try:
+            vlm = VLM(vlm.lower())
+        except ValueError:
+            raise ValueError(
+                f"Invalid vlm value: {vlm}. Must be one of {[e.value for e in VLM]}"
+            )
+
+    if not isinstance(result, RESULT_TYPES[vlm]):
+        raise ValueError(
+            f"Invalid VLM result type: {type(result)}. Must be {RESULT_TYPES[vlm]}"
+        )
+
+    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(vlm, [])
+    for arg in kwargs:
+        if arg not in allowed_args:
+            raise ValueError(f"Argument {arg} is not allowed for {vlm.name}")
+
+    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], 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[
+    np.ndarray, Optional[np.ndarray], Optional[np.ndarray], Optional[np.ndarray]
+]:
+    """
+    Parse results from the Florence 2 multi-model model.
+    https://huggingface.co/microsoft/Florence-2-large
+
+    Parameters:
+        result: dict containing the model output
+
+    Returns:
+        xyxy (np.ndarray): An array of shape `(n, 4)` containing
+            the bounding boxes coordinates in format `[x1, y1, x2, y2]`
+        labels: (Optional[np.ndarray]): An array of shape `(n,)` containing
+            the class labels for each bounding box
+        masks: (Optional[np.ndarray]): An array of shape `(n, h, w)` containing
+            the segmentation masks for each bounding box
+        obb_boxes: (Optional[np.ndarray]): An array of shape `(n, 4, 2)` containing
+            oriented bounding boxes.
+    """
+    assert len(result) == 1, f"Expected result with a single element. Got: {result}"
+    task = next(iter(result.keys()))
+    if task not in SUPPORTED_TASKS_FLORENCE_2:
+        raise ValueError(
+            f"{task} not supported. Supported tasks are: {SUPPORTED_TASKS_FLORENCE_2}"
+        )
+    result = result[task]
+
+    if task in ["<OD>", "<CAPTION_TO_PHRASE_GROUNDING>", "<DENSE_REGION_CAPTION>"]:
+        xyxy = np.array(result["bboxes"], dtype=np.float32)
+        labels = np.array(result["labels"])
+        return xyxy, labels, None, None
+
+    if task == "<REGION_PROPOSAL>":
+        xyxy = np.array(result["bboxes"], dtype=np.float32)
+        # provides labels, but they are ["", "", "", ...]
+        return xyxy, None, None, None
+
+    if task == "<OCR_WITH_REGION>":
+        xyxyxyxy = np.array(result["quad_boxes"], dtype=np.float32)
+        xyxyxyxy = xyxyxyxy.reshape(-1, 4, 2)
+        xyxy = np.array([polygon_to_xyxy(polygon) for polygon in xyxyxyxy])
+        labels = np.array(result["labels"])
+        return xyxy, labels, None, xyxyxyxy
+
+    if task in ["<REFERRING_EXPRESSION_SEGMENTATION>", "<REGION_TO_SEGMENTATION>"]:
+        xyxy_list = []
+        masks_list = []
+        for polygons_of_same_class in result["polygons"]:
+            for polygon in polygons_of_same_class:
+                polygon = np.reshape(polygon, (-1, 2)).astype(np.int32)
+                mask = polygon_to_mask(polygon, resolution_wh).astype(bool)
+                masks_list.append(mask)
+                xyxy = polygon_to_xyxy(polygon)
+                xyxy_list.append(xyxy)
+            # per-class labels also provided, but they are ["", "", "", ...]
+            # when we figure out how to set class names, we can do
+            # zip(result["labels"], result["polygons"])
+        xyxy = np.array(xyxy_list, dtype=np.float32)
+        masks = np.array(masks_list)
+        return xyxy, None, masks, None
+
+    if task == "<OPEN_VOCABULARY_DETECTION>":
+        xyxy = np.array(result["bboxes"], dtype=np.float32)
+        labels = np.array(result["bboxes_labels"])
+        # Also has "polygons" and "polygons_labels", but they don't seem to be used
+        return xyxy, labels, None, None
+
+    if task in ["<REGION_TO_CATEGORY>", "<REGION_TO_DESCRIPTION>"]:
+        assert isinstance(result, str), (
+            f"Expected string as <REGION_TO_CATEGORY> result, got {type(result)}"
+        )
+
+        if result == "No object detected.":
+            return np.empty((0, 4), dtype=np.float32), np.array([]), None, None
+
+        pattern = re.compile(r"<loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)>")
+        match = pattern.search(result)
+        assert match is not None, (
+            f"Expected string to end in location tags, but got {result}"
+        )
+
+        w, h = resolution_wh
+        xyxy = np.array([match.groups()], dtype=np.float32)
+        xyxy *= np.array([w, h, w, h]) / 1000
+        result_string = result[: match.start()]
+        labels = np.array([result_string])
+        return xyxy, labels, None, None
+
+    assert False, f"Unimplemented task: {task}"

supervision/draw/color.py

@@ -83,15 +83,16 @@ class Color:
         # Color(r=255, g=255, b=255)
         ```
 
-    | Constant   | Hex Code   | RGB              |
-    |------------|------------|------------------|
-    | `WHITE`    | `#FFFFFF`  | `(255, 255, 255)`|
-    | `BLACK`    | `#000000`  | `(0, 0, 0)`      |
-    | `RED`      | `#FF0000`  | `(255, 0, 0)`    |
-    | `GREEN`    | `#00FF00`  | `(0, 255, 0)`    |
-    | `BLUE`     | `#0000FF`  | `(0, 0, 255)`    |
-    | `YELLOW`   | `#FFFF00`  | `(255, 255, 0)`  |
-    | `ROBOFLOW` | `#A351FB`  | `(163, 81, 251)` |
+    | Constant   | Hex Code   | RGB               |
+    |------------|------------|-------------------|
+    | `WHITE`    | `#FFFFFF`  | `(255, 255, 255)` |
+    | `BLACK`    | `#000000`  | `(0, 0, 0)`       |
+    | `GREY`     | `#808080`  | `(128, 128, 128)` |
+    | `RED`      | `#FF0000`  | `(255, 0, 0)`     |
+    | `GREEN`    | `#00FF00`  | `(0, 255, 0)`     |
+    | `BLUE`     | `#0000FF`  | `(0, 0, 255)`     |
+    | `YELLOW`   | `#FFFF00`  | `(255, 255, 0)`   |
+    | `ROBOFLOW` | `#A351FB`  | `(163, 81, 251)`  |
     """
 
     r: int
@@ -235,6 +236,10 @@ class Color:
     def BLACK(cls) -> Color:
         return Color.from_hex("#000000")
 
+    @classproperty
+    def GREY(cls) -> Color:
+        return Color.from_hex("#808080")
+
     @classproperty
     def RED(cls) -> Color:
         return Color.from_hex("#FF0000")

supervision/keypoint/core.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-from contextlib import suppress
 from dataclasses import dataclass, field
 from typing import Any, Dict, Iterable, Iterator, List, Optional, Tuple, Union
 
@@ -24,8 +23,8 @@ class KeyPoints:
     === "Ultralytics"
 
         Use [`sv.KeyPoints.from_ultralytics`](/latest/keypoint/core/#supervision.keypoint.core.KeyPoints.from_ultralytics)
-        method, which accepts [YOLOv8](https://github.com/ultralytics/ultralytics)
-        pose result.
+        method, which accepts [YOLOv8-pose](https://docs.ultralytics.com/models/yolov8/), [YOLO11-pose](https://docs.ultralytics.com/models/yolo11/)
+        [pose](https://docs.ultralytics.com/tasks/pose/) result.
 
         ```python
         import cv2
@@ -33,7 +32,7 @@ class KeyPoints:
         from ultralytics import YOLO
 
         image = cv2.imread(<SOURCE_IMAGE_PATH>)
-        model = YOLO('yolov8s-pose.pt')
+        model = YOLO('yolo11s-pose.pt')
 
         result = model(image)[0]
         key_points = sv.KeyPoints.from_ultralytics(result)
@@ -62,6 +61,7 @@ class KeyPoints:
         method, which accepts [MediaPipe](https://github.com/google-ai-edge/mediapipe)
         pose result.
 
+
         ```python
         import cv2
         import mediapipe as mp
@@ -92,10 +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`
@@ -204,9 +204,10 @@ class KeyPoints:
                 "You can retrieve it like so:  inference_result = model.infer(image)[0]"
             )
 
-        with suppress(AttributeError):
+        if hasattr(inference_result, "dict"):
             inference_result = inference_result.dict(exclude_none=True, by_alias=True)
-
+        elif hasattr(inference_result, "json"):
+            inference_result = inference_result.json()
         if not inference_result.get("predictions"):
             return cls.empty()
 
@@ -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/__init__.py

@@ -19,3 +19,22 @@ from supervision.metrics.utils.object_size import (
     get_detection_size_category,
     get_object_size_category,
 )
+
+__all__ = [
+    "AveragingMethod",
+    "F1Score",
+    "F1ScoreResult",
+    "MeanAveragePrecision",
+    "MeanAveragePrecisionResult",
+    "MeanAverageRecall",
+    "MeanAverageRecallResult",
+    "Metric",
+    "MetricTarget",
+    "ObjectSizeCategory",
+    "Precision",
+    "PrecisionResult",
+    "Recall",
+    "RecallResult",
+    "get_detection_size_category",
+    "get_object_size_category",
+]

supervision/metrics/mean_average_recall.py

@@ -670,8 +670,7 @@ class MeanAverageRecallResult:
         ax.set_ylim(0, 1)
         ax.set_ylabel("Value", fontweight="bold")
         title = (
-            f"Mean Average Recall, by Object Size"
-            f"\n(target: {self.metric_target.value})"
+            f"Mean Average Recall, by Object Size\n(target: {self.metric_target.value})"
         )
         ax.set_title(title, fontweight="bold")
 

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]'`"
         )

supervision/utils/video.py

@@ -7,6 +7,7 @@ from typing import Callable, Generator, Optional, Tuple
 
 import cv2
 import numpy as np
+from tqdm.auto import tqdm
 
 
 @dataclass
@@ -195,6 +196,9 @@ def process_video(
     source_path: str,
     target_path: str,
     callback: Callable[[np.ndarray, int], np.ndarray],
+    max_frames: Optional[int] = None,
+    show_progress: bool = False,
+    progress_message: str = "Processing video",
 ) -> None:
     """
     Process a video file by applying a callback function on each frame
@@ -207,6 +211,9 @@ def process_video(
             a numpy ndarray representation of a video frame and an
             int index of the frame and returns a processed numpy ndarray
             representation of the frame.
+        max_frames (Optional[int]): The maximum number of frames to process.
+        show_progress (bool): Whether to show a progress bar.
+        progress_message (str): The message to display in the progress bar.
 
     Examples:
         ```python
@@ -223,12 +230,29 @@ def process_video(
         ```
     """
     source_video_info = VideoInfo.from_video_path(video_path=source_path)
+    video_frames_generator = get_video_frames_generator(
+        source_path=source_path, end=max_frames
+    )
     with VideoSink(target_path=target_path, video_info=source_video_info) as sink:
+        total_frames = (
+            min(source_video_info.total_frames, max_frames)
+            if max_frames is not None
+            else source_video_info.total_frames
+        )
         for index, frame in enumerate(
-            get_video_frames_generator(source_path=source_path)
+            tqdm(
+                video_frames_generator,
+                total=total_frames,
+                disable=not show_progress,
+                desc=progress_message,
+            )
         ):
             result_frame = callback(frame, index)
             sink.write_frame(frame=result_frame)
+        else:
+            for index, frame in enumerate(video_frames_generator):
+                result_frame = callback(frame, index)
+                sink.write_frame(frame=result_frame)
 
 
 class FPSMonitor:

supervision/validators/__init__.py

@@ -54,7 +54,7 @@ def validate_confidence(confidence: Any, n: int) -> None:
 
 
 def validate_keypoint_confidence(confidence: Any, n: int, m: int) -> None:
-    expected_shape = f"({n,m})"
+    expected_shape = f"({n, m})"
     actual_shape = str(getattr(confidence, "shape", None))
 
     if confidence is not None:

test/detection/test_core.py

@@ -223,6 +223,12 @@ TEST_DET_DIFFERENT_METADATA = Detections(
             None,
             pytest.raises(IndexError),
         ),
+        (
+            Detections.empty(),
+            np.isin(Detections.empty()["class_name"], ["cat", "dog"]),
+            Detections.empty(),
+            DoesNotRaise(),
+        ),  # Filter an empty detections by specific class names
     ],
 )
 def test_getitem(

test/detection/test_csv.py

@@ -408,8 +408,8 @@ def assert_csv_equal(file_name, expected_rows):
     with open(file_name, mode="r", newline="") as file:
         reader = csv.reader(file)
         for i, row in enumerate(reader):
-            assert (
-                [str(item) for item in expected_rows[i]] == row
-            ), f"Row in CSV didn't match expected output: {row} != {expected_rows[i]}"
+            assert [str(item) for item in expected_rows[i]] == row, (
+                f"Row in CSV didn't match expected output: {row} != {expected_rows[i]}"
+            )
 
     os.remove(file_name)

test/detection/test_json.py

@@ -242,8 +242,8 @@ def test_json_sink(
 def assert_json_equal(file_name, expected_rows):
     with open(file_name, "r") as file:
         data = json.load(file)
-        assert (
-            data == expected_rows
-        ), f"Data in JSON file didn't match expected output: {data} != {expected_rows}"
+        assert data == expected_rows, (
+            f"Data in JSON file didn't match expected output: {data} != {expected_rows}"
+        )
 
     os.remove(file_name)

test/detection/test_line_counter.py

@@ -250,12 +250,12 @@ def test_line_zone_one_detection_default_anchors(
         crossed_in_list.append(crossed_in[0])
         crossed_out_list.append(crossed_out[0])
 
-    assert (
-        crossed_in_list == expected_crossed_in
-    ), f"expected {expected_crossed_in}, got {crossed_in_list}"
-    assert (
-        crossed_out_list == expected_crossed_out
-    ), f"expected {expected_crossed_out}, got {crossed_out_list}"
+    assert crossed_in_list == expected_crossed_in, (
+        f"expected {expected_crossed_in}, got {crossed_in_list}"
+    )
+    assert crossed_out_list == expected_crossed_out, (
+        f"expected {expected_crossed_out}, got {crossed_out_list}"
+    )
 
 
 @pytest.mark.parametrize(
@@ -415,12 +415,12 @@ def test_line_zone_one_detection(
         crossed_in_list.append(crossed_in[0])
         crossed_out_list.append(crossed_out[0])
 
-    assert (
-        crossed_in_list == expected_crossed_in
-    ), f"expected {expected_crossed_in}, got {crossed_in_list}"
-    assert (
-        crossed_out_list == expected_crossed_out
-    ), f"expected {expected_crossed_out}, got {crossed_out_list}"
+    assert crossed_in_list == expected_crossed_in, (
+        f"expected {expected_crossed_in}, got {crossed_in_list}"
+    )
+    assert crossed_out_list == expected_crossed_out, (
+        f"expected {expected_crossed_out}, got {crossed_out_list}"
+    )
 
 
 @pytest.mark.parametrize(
@@ -600,12 +600,12 @@ def test_line_zone_one_detection_long_horizon(
         crossed_in_list.append(crossed_in[0])
         crossed_out_list.append(crossed_out[0])
 
-    assert (
-        crossed_in_list == expected_crossed_in
-    ), f"expected {expected_crossed_in}, got {crossed_in_list}"
-    assert (
-        crossed_out_list == expected_crossed_out
-    ), f"expected {expected_crossed_out}, got {crossed_out_list}"
+    assert crossed_in_list == expected_crossed_in, (
+        f"expected {expected_crossed_in}, got {crossed_in_list}"
+    )
+    assert crossed_out_list == expected_crossed_out, (
+        f"expected {expected_crossed_out}, got {crossed_out_list}"
+    )
 
 
 @pytest.mark.parametrize(

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

@@ -16,10 +16,13 @@ from supervision.detection.utils import (
     merge_data,
     merge_metadata,
     move_boxes,
+    move_masks,
     process_roboflow_result,
     scale_boxes,
     xcycwh_to_xyxy,
     xywh_to_xyxy,
+    xyxy_to_xcycarh,
+    xyxy_to_xywh,
 )
 
 TEST_MASK = np.zeros((1, 1000, 1000), dtype=bool)
@@ -387,13 +390,13 @@ def test_process_roboflow_result(
         )
         for key in result[5]:
             if isinstance(result[5][key], np.ndarray):
-                assert np.array_equal(
-                    result[5][key], expected_result[5][key]
-                ), f"Mismatch in arrays for key {key}"
+                assert np.array_equal(result[5][key], expected_result[5][key]), (
+                    f"Mismatch in arrays for key {key}"
+                )
             else:
-                assert (
-                    result[5][key] == expected_result[5][key]
-                ), f"Mismatch in non-array data for key {key}"
+                assert result[5][key] == expected_result[5][key], (
+                    f"Mismatch in non-array data for key {key}"
+                )
 
 
 @pytest.mark.parametrize(
@@ -442,6 +445,268 @@ def test_move_boxes(
         assert np.array_equal(result, expected_result)
 
 
+@pytest.mark.parametrize(
+    "masks, offset, resolution_wh, expected_result, exception",
+    [
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([0, 0]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([-1, -1]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [True, True, False, False],
+                        [True, True, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([-2, -2]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [True, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([-3, -3]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([-2, -1]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [True, False, False, False],
+                        [True, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([-1, -2]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [True, True, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([-2, 2]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [True, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([3, 3]),
+            (4, 4),
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+        (
+            np.array(
+                [
+                    [
+                        [False, False, False, False],
+                        [False, True, True, False],
+                        [False, True, True, False],
+                        [False, False, False, False],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            np.array([3, 3]),
+            (6, 6),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False, False],
+                        [False, False, False, False, False, False],
+                        [False, False, False, False, False, False],
+                        [False, False, False, False, False, False],
+                        [False, False, False, False, True, True],
+                        [False, False, False, False, True, True],
+                    ]
+                ],
+                dtype=bool,
+            ),
+            DoesNotRaise(),
+        ),
+    ],
+)
+def test_move_masks(
+    masks: np.ndarray,
+    offset: np.ndarray,
+    resolution_wh: Tuple[int, int],
+    expected_result: np.ndarray,
+    exception: Exception,
+) -> None:
+    with exception:
+        result = move_masks(masks=masks, offset=offset, resolution_wh=resolution_wh)
+        np.testing.assert_array_equal(result, expected_result)
+
+
 @pytest.mark.parametrize(
     "xyxy, factor, expected_result, exception",
     [
@@ -778,13 +1043,13 @@ def test_merge_data(
 
         for key in result:
             if isinstance(result[key], np.ndarray):
-                assert np.array_equal(
-                    result[key], expected_result[key]
-                ), f"Mismatch in arrays for key {key}"
+                assert np.array_equal(result[key], expected_result[key]), (
+                    f"Mismatch in arrays for key {key}"
+                )
             else:
-                assert (
-                    result[key] == expected_result[key]
-                ), f"Mismatch in non-array data for key {key}"
+                assert result[key] == expected_result[key], (
+                    f"Mismatch in non-array data for key {key}"
+                )
 
 
 @pytest.mark.parametrize(
@@ -951,13 +1216,13 @@ def test_get_data_item(
         result = get_data_item(data=data, index=index)
         for key in result:
             if isinstance(result[key], np.ndarray):
-                assert np.array_equal(
-                    result[key], expected_result[key]
-                ), f"Mismatch in arrays for key {key}"
+                assert np.array_equal(result[key], expected_result[key]), (
+                    f"Mismatch in arrays for key {key}"
+                )
             else:
-                assert (
-                    result[key] == expected_result[key]
-                ), f"Mismatch in non-array data for key {key}"
+                assert result[key] == expected_result[key], (
+                    f"Mismatch in non-array data for key {key}"
+                )
 
 
 @pytest.mark.parametrize(
@@ -1118,6 +1383,80 @@ 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(
+    "xyxy, expected_result",
+    [
+        # Empty and zero cases
+        (np.array([]).reshape(0, 4), np.array([]).reshape(0, 4)),  # empty array
+        (
+            np.array([[0, 0, 0, 0]]),
+            np.array([[0, 0, 0.0, 0]]),
+        ),  # zero size bounding box
+        (
+            np.array([[10, 10, 10, 10]]),
+            np.array([[10, 10, 0.0, 0]]),
+        ),  # point (x1=x2, y1=y2)
+        # Zero width/height cases
+        (np.array([[50, 50, 80, 50]]), np.array([[65, 50, 0.0, 0]])),  # zero height
+        (np.array([[50, 50, 50, 80]]), np.array([[50, 65, 0.0, 30]])),  # zero width
+        # Standard cases
+        (np.array([[10, 20, 40, 60]]), np.array([[25, 40, 0.75, 40]])),  # standard case
+        (
+            np.array([[-30, -40, -10, -20]]),
+            np.array([[-20, -30, 1.0, 20]]),
+        ),  # all negative values
+        (
+            np.array([[0.1, 0.2, 0.4, 0.6]]),
+            np.array([[0.25, 0.4, 0.75, 0.4]]),
+        ),  # values between 0-1
+        # Different aspect ratios
+        (
+            np.array([[10, 20, 50, 100]]),
+            np.array([[30, 60, 0.5, 80]]),
+        ),  # tall rectangle (height > width)
+        (
+            np.array([[20, 10, 100, 50]]),
+            np.array([[60, 30, 2.0, 40]]),
+        ),  # wide rectangle (width > height)
+        (
+            np.array([[50, 50, 150, 150]]),
+            np.array([[100, 100, 1.0, 100]]),
+        ),  # height == width
+        # Multiple boxes in one array
+        (
+            np.array([[0, 0, 0, 0], [10, 20, 40, 60]]),
+            np.array([[0, 0, 0.0, 0], [25, 40, 0.75, 40]]),
+        ),  # one zero-sized box and one normal box
+    ],
+)
+def test_xyxy_to_xcycarh(xyxy: np.ndarray, expected_result: np.ndarray) -> None:
+    result = xyxy_to_xcycarh(xyxy)
+    np.testing.assert_allclose(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(

test/detection/tools/test_inference_slicer.py

@@ -22,33 +22,26 @@ def mock_callback():
 @pytest.mark.parametrize(
     "slice_wh, overlap_ratio_wh, overlap_wh, expected_overlap, exception",
     [
-        # Valid case: overlap_ratio_wh provided, overlap calculated from the ratio
-        ((128, 128), (0.2, 0.2), None, None, DoesNotRaise()),
-        # Valid case: overlap_wh in pixels, no ratio provided
+        # Valid case: explicit overlap_wh in pixels
+        ((128, 128), None, (26, 26), (26, 26), DoesNotRaise()),
+        # Valid case: overlap_wh in pixels
         ((128, 128), None, (20, 20), (20, 20), DoesNotRaise()),
-        # Invalid case: overlap_ratio_wh greater than 1, should raise ValueError
-        ((128, 128), (1.1, 0.5), None, None, pytest.raises(ValueError)),
         # Invalid case: negative overlap_wh, should raise ValueError
         ((128, 128), None, (-10, 20), None, pytest.raises(ValueError)),
-        # Invalid case:
-        # overlap_ratio_wh and overlap_wh provided, should raise ValueError
-        ((128, 128), (0.5, 0.5), (20, 20), (20, 20), pytest.raises(ValueError)),
-        # Valid case: no overlap_ratio_wh, overlap_wh = 50 pixels
-        ((256, 256), None, (50, 50), (50, 50), DoesNotRaise()),
-        # Valid case: overlap_ratio_wh provided, overlap calculated from (0.3, 0.3)
-        ((200, 200), (0.3, 0.3), None, None, DoesNotRaise()),
-        # Valid case: small overlap_ratio_wh values
-        ((100, 100), (0.1, 0.1), None, None, DoesNotRaise()),
-        # Invalid case: negative overlap_ratio_wh value, should raise ValueError
-        ((128, 128), (-0.1, 0.2), None, None, pytest.raises(ValueError)),
-        # Invalid case: negative overlap_ratio_wh with overlap_wh provided
-        ((128, 128), (-0.1, 0.2), (30, 30), None, pytest.raises(ValueError)),
-        # Invalid case: overlap_wh greater than slice size, should raise ValueError
-        ((128, 128), None, (150, 150), (150, 150), DoesNotRaise()),
-        # Valid case: overlap_ratio_wh is 0, no overlap
-        ((128, 128), (0.0, 0.0), None, None, DoesNotRaise()),
-        # Invalid case: no overlaps defined, no overlap
+        # Invalid case: no overlaps defined
         ((128, 128), None, None, None, pytest.raises(ValueError)),
+        # Valid case: overlap_wh = 50 pixels
+        ((256, 256), None, (50, 50), (50, 50), DoesNotRaise()),
+        # Valid case: overlap_wh = 60 pixels
+        ((200, 200), None, (60, 60), (60, 60), DoesNotRaise()),
+        # Valid case: small overlap_wh values
+        ((100, 100), None, (0.1, 0.1), (0.1, 0.1), DoesNotRaise()),
+        # Invalid case: negative overlap_wh values
+        ((128, 128), None, (-10, -10), None, pytest.raises(ValueError)),
+        # Invalid case: overlap_wh greater than slice size
+        ((128, 128), None, (150, 150), (150, 150), DoesNotRaise()),
+        # Valid case: zero overlap
+        ((128, 128), None, (0, 0), (0, 0), DoesNotRaise()),
     ],
 )
 def test_inference_slicer_overlap(
@@ -187,6 +180,6 @@ def test_generate_offset(
     )
 
     # Verify that the generated offsets match the expected offsets
-    assert np.array_equal(
-        offsets, expected_offsets
-    ), f"Expected {expected_offsets}, got {offsets}"
+    assert np.array_equal(offsets, expected_offsets), (
+        f"Expected {expected_offsets}, got {offsets}"
+    )

test/utils/test_conversion.py

@@ -22,15 +22,15 @@ def test_ensure_cv2_image_for_processing_when_pillow_image_submitted(
         param_a: int,
         param_b: str,
     ) -> np.ndarray:
-        assert np.allclose(
-            image, empty_cv2_image
-        ), "Expected conversion to OpenCV image to happen"
-        assert (
-            param_a == param_a_value
-        ), f"Parameter a expected to be {param_a_value} in target function"
-        assert (
-            param_b == param_b_value
-        ), f"Parameter b expected to be {param_b_value} in target function"
+        assert np.allclose(image, empty_cv2_image), (
+            "Expected conversion to OpenCV image to happen"
+        )
+        assert param_a == param_a_value, (
+            f"Parameter a expected to be {param_a_value} in target function"
+        )
+        assert param_b == param_b_value, (
+            f"Parameter b expected to be {param_b_value} in target function"
+        )
         return image
 
     # when
@@ -61,15 +61,15 @@ def test_ensure_cv2_image_for_processing_when_cv2_image_submitted(
         param_a: int,
         param_b: str,
     ) -> np.ndarray:
-        assert np.allclose(
-            image, empty_cv2_image
-        ), "Expected conversion to OpenCV image to happen"
-        assert (
-            param_a == param_a_value
-        ), f"Parameter a expected to be {param_a_value} in target function"
-        assert (
-            param_b == param_b_value
-        ), f"Parameter b expected to be {param_b_value} in target function"
+        assert np.allclose(image, empty_cv2_image), (
+            "Expected conversion to OpenCV image to happen"
+        )
+        assert param_a == param_a_value, (
+            f"Parameter a expected to be {param_a_value} in target function"
+        )
+        assert param_b == param_b_value, (
+            f"Parameter b expected to be {param_b_value} in target function"
+        )
         return image
 
     # when
@@ -91,9 +91,9 @@ def test_cv2_to_pillow(
 
     # then
     difference = ImageChops.difference(result, empty_pillow_image)
-    assert (
-        difference.getbbox() is None
-    ), "Conversion to PIL.Image expected not to change the content of image"
+    assert difference.getbbox() is None, (
+        "Conversion to PIL.Image expected not to change the content of image"
+    )
 
 
 def test_pillow_to_cv2(
@@ -103,9 +103,9 @@ def test_pillow_to_cv2(
     result = pillow_to_cv2(image=empty_pillow_image)
 
     # then
-    assert np.allclose(
-        result, empty_cv2_image
-    ), "Conversion to OpenCV image expected not to change the content of image"
+    assert np.allclose(result, empty_cv2_image), (
+        "Conversion to OpenCV image expected not to change the content of image"
+    )
 
 
 def test_images_to_cv2_when_empty_input_provided() -> None:
@@ -128,9 +128,9 @@ def test_images_to_cv2_when_only_cv2_images_provided(
     # then
     assert len(result) == 5, "Expected the same number of output element as input ones"
     for result_element in result:
-        assert (
-            result_element is empty_cv2_image
-        ), "Expected CV images not to be touched by conversion"
+        assert result_element is empty_cv2_image, (
+            "Expected CV images not to be touched by conversion"
+        )
 
 
 def test_images_to_cv2_when_only_pillow_images_provided(
@@ -146,9 +146,9 @@ def test_images_to_cv2_when_only_pillow_images_provided(
     # then
     assert len(result) == 5, "Expected the same number of output element as input ones"
     for result_element in result:
-        assert np.allclose(
-            result_element, empty_cv2_image
-        ), "Output images expected to be equal to empty OpenCV image"
+        assert np.allclose(result_element, empty_cv2_image), (
+            "Output images expected to be equal to empty OpenCV image"
+        )
 
 
 def test_images_to_cv2_when_mixed_input_provided(
@@ -163,9 +163,9 @@ def test_images_to_cv2_when_mixed_input_provided(
 
     # then
     assert len(result) == 2, "Expected the same number of output element as input ones"
-    assert np.allclose(
-        result[0], empty_cv2_image
-    ), "PIL image should be converted to OpenCV one, equal to example empty image"
-    assert (
-        result[1] is empty_cv2_image
-    ), "Expected CV images not to be touched by conversion"
+    assert np.allclose(result[0], empty_cv2_image), (
+        "PIL image should be converted to OpenCV one, equal to example empty image"
+    )
+    assert result[1] is empty_cv2_image, (
+        "Expected CV images not to be touched by conversion"
+    )

test/utils/test_image.py

@@ -21,9 +21,9 @@ def test_resize_image_for_opencv_image() -> None:
     )
 
     # then
-    assert np.allclose(
-        result, expected_result
-    ), "Expected output shape to be (w, h): (1024, 768)"
+    assert np.allclose(result, expected_result), (
+        "Expected output shape to be (w, h): (1024, 768)"
+    )
 
 
 def test_resize_image_for_pillow_image() -> None:
@@ -41,9 +41,9 @@ def test_resize_image_for_pillow_image() -> None:
     # then
     assert result.size == (1024, 768), "Expected output shape to be (w, h): (1024, 768)"
     difference = ImageChops.difference(result, expected_result)
-    assert (
-        difference.getbbox() is None
-    ), "Expected no difference in resized image content as the image is all zeros"
+    assert difference.getbbox() is None, (
+        "Expected no difference in resized image content as the image is all zeros"
+    )
 
 
 def test_letterbox_image_for_opencv_image() -> None:
@@ -95,9 +95,9 @@ def test_letterbox_image_for_pillow_image() -> None:
         1024,
     ), "Expected output shape to be (w, h): (1024, 1024)"
     difference = ImageChops.difference(result, expected_result)
-    assert (
-        difference.getbbox() is None
-    ), "Expected padding to be added top and bottom with padding added top and bottom"
+    assert difference.getbbox() is None, (
+        "Expected padding to be added top and bottom with padding added top and bottom"
+    )
 
 
 def test_create_tiles_with_one_image(