diff --git a/.github/workflows/codespell.yml b/.github/workflows/codespell.yml index 563f51df82..5c61d13720 100644 --- a/.github/workflows/codespell.yml +++ b/.github/workflows/codespell.yml @@ -15,7 +15,7 @@ jobs: steps: - name: Checkout - uses: actions/checkout@v3 + uses: actions/checkout@v6 - name: Annotate locations with typos uses: codespell-project/codespell-problem-matcher@v1 - name: Codespell diff --git a/.github/workflows/docs_and_notebooks_checks.yml b/.github/workflows/docs_and_notebooks_checks.yml new file mode 100644 index 0000000000..b35b68eda7 --- /dev/null +++ b/.github/workflows/docs_and_notebooks_checks.yml @@ -0,0 +1,59 @@ +name: Docs & notebooks freshness and formatting checks + +on: + pull_request: + branches: [main] + push: + branches: [main] + +permissions: + contents: read + +jobs: + staleness: + name: Docs and notebooks scan (read-only) + runs-on: ubuntu-latest + timeout-minutes: 5 + + steps: + - name: Checkout repository (full history for git dates) + uses: actions/checkout@v6 + with: + fetch-depth: 0 + + - name: Set up Python + uses: actions/setup-python@v6 + with: + python-version: "3.12" + + - name: Install staleness tool dependencies + run: | + python -m pip install --upgrade pip + python -m pip install "pydantic>=2,<3" pyyaml "nbformat>=5" + + - name: Run staleness report (read-only) + run: | + python tools/docs_and_notebooks_check.py \ + --config tools/docs_and_notebooks_report_config.yml \ + --out-dir tmp/docs_nb_checks \ + report + + + # Optional: run check mode (will fail only once you populate allowlists in config) + - name: Run staleness policy check (optional gate) + continue-on-error: true + run: | + python tools/docs_and_notebooks_check.py \ + --config tools/docs_and_notebooks_report_config.yml \ + --out-dir tmp/docs_nb_checks \ + --no-step-summary \ + check + + - name: Upload staleness artifacts + uses: actions/upload-artifact@v4 + with: + name: staleness-report + path: | + tmp/docs_nb_checks/*.json + tmp/docs_nb_checks/*.md + if-no-files-found: error diff --git a/.github/workflows/python-package.yml b/.github/workflows/python-package.yml index 71d5df0efe..c635d034d6 100644 --- a/.github/workflows/python-package.yml +++ b/.github/workflows/python-package.yml @@ -87,7 +87,7 @@ jobs: shell: bash -el {0} # Important: activates the conda environment run: | python -m pip install --upgrade pip setuptools wheel - pip install --no-cache-dir -e . + pip install --no-cache-dir -e . --group dev - name: Install ffmpeg run: | diff --git a/.gitignore b/.gitignore index 9d2c3a6efe..30b2b14bff 100644 --- a/.gitignore +++ b/.gitignore @@ -136,6 +136,8 @@ ENV/ # mypy .mypy_cache/ +# Automated docs checks +**/tmp/docs_nb_checks/ # Automatic test selection report diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index f1d96da46b..551e3f7dda 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -26,3 +26,20 @@ repos: hooks: - id: black language_version: python3 + - repo: local + hooks: + - id: dlc-docs-notebooks-check + name: DLC docs+notebooks staleness/check + nbformat validate + normalization + entry: python tools/docs_and_notebooks_check.py + language: python + pass_filenames: true + files: ^(docs/|examples/(JUPYTER|COLAB)/|tools/).*(\.md|\.ipynb)$ + args: + - --config + - tools/docs_and_notebooks_report_config.yml + - check + - --targets + additional_dependencies: + - "pydantic>=2,<3" + - "pyyaml" + - "nbformat>=5" diff --git a/docs/Governance.md b/docs/Governance.md index 187379cc1c..2ee2b1b172 100644 --- a/docs/Governance.md +++ b/docs/Governance.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- (governance-model)= # Governance Model of DeepLabCut (adapted from https://napari.org/stable/community/governance.html) diff --git a/docs/HelperFunctions.md b/docs/HelperFunctions.md index e4a5aeebe8..25efc177fa 100644 --- a/docs/HelperFunctions.md +++ b/docs/HelperFunctions.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- (helper-functions)= # Helper & Advanced Optional Function Documentation diff --git a/docs/MISSION_AND_VALUES.md b/docs/MISSION_AND_VALUES.md index c82c4b7a39..bc6623a6af 100644 --- a/docs/MISSION_AND_VALUES.md +++ b/docs/MISSION_AND_VALUES.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- (mission-and-values)= # Mission and Values of DeepLabCut diff --git a/docs/ModelZoo.md b/docs/ModelZoo.md index 4ae7c8dbaa..9d37486afc 100644 --- a/docs/ModelZoo.md +++ b/docs/ModelZoo.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-07-06' + last_metadata_updated: '2026-03-06' + ignore: false +--- (file:model-zoo)= # The DeepLabCut Model Zoo! diff --git a/docs/Overviewof3D.md b/docs/Overviewof3D.md index 6a83d7575e..75b498a686 100644 --- a/docs/Overviewof3D.md +++ b/docs/Overviewof3D.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-10-14' + last_metadata_updated: '2026-03-06' + ignore: false +--- (3D-overview)= # 3D DeepLabCut diff --git a/docs/README.md b/docs/README.md index df606c21b6..1114741159 100644 --- a/docs/README.md +++ b/docs/README.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2022-08-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- Please see https://deeplabcut.github.io/DeepLabCut for documentation on how to use this software. This directory contains the source code for the docs. diff --git a/docs/UseOverviewGuide.md b/docs/UseOverviewGuide.md index 017b30dcc4..e27c462310 100644 --- a/docs/UseOverviewGuide.md +++ b/docs/UseOverviewGuide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- (overview)= # 🥳 Get started with DeepLabCut: our key recommendations diff --git a/docs/beginner-guides/Training-Evaluation.md b/docs/beginner-guides/Training-Evaluation.md index 18ab3bf736..2ec7a9157a 100644 --- a/docs/beginner-guides/Training-Evaluation.md +++ b/docs/beginner-guides/Training-Evaluation.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Neural Network training and evaluation in the GUI DLC LIVE! diff --git a/docs/beginner-guides/beginners-guide.md b/docs/beginner-guides/beginners-guide.md index f3cfea194d..e4e466b8a2 100644 --- a/docs/beginner-guides/beginners-guide.md +++ b/docs/beginner-guides/beginners-guide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-03-03' + last_metadata_updated: '2026-03-06' + ignore: false +--- (beginners-guide)= # Using DeepLabCut DLC LIVE! diff --git a/docs/beginner-guides/labeling.md b/docs/beginner-guides/labeling.md index f3a0ec10d3..3c79807a3a 100644 --- a/docs/beginner-guides/labeling.md +++ b/docs/beginner-guides/labeling.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- (labeling)= # Labeling GUI diff --git a/docs/beginner-guides/manage-project.md b/docs/beginner-guides/manage-project.md index ff32aef0e2..ee0e0f339e 100644 --- a/docs/beginner-guides/manage-project.md +++ b/docs/beginner-guides/manage-project.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Setting up what keypoints to track DLC LIVE! diff --git a/docs/beginner-guides/video-analysis.md b/docs/beginner-guides/video-analysis.md index 126c7bdf9a..132feeb812 100644 --- a/docs/beginner-guides/video-analysis.md +++ b/docs/beginner-guides/video-analysis.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Video Analysis with DeepLabCut DLC LIVE! diff --git a/docs/benchmark.md b/docs/benchmark.md index 612e307d0f..1b2e9a5d42 100644 --- a/docs/benchmark.md +++ b/docs/benchmark.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # DeepLabCut benchmark For further information and the leaderboard, see [the official homepage](https://benchmark.deeplabcut.org/). diff --git a/docs/citation.md b/docs/citation.md index c427b1e223..635bd62662 100644 --- a/docs/citation.md +++ b/docs/citation.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2024-10-27' + last_metadata_updated: '2026-03-06' + ignore: false +--- # How to Cite DeepLabCut Thank you for using DeepLabCut! Here are our recommendations for citing and documenting your use of DeepLabCut in your Methods section: diff --git a/docs/convert_maDLC.md b/docs/convert_maDLC.md index 19dd017692..14e697c719 100644 --- a/docs/convert_maDLC.md +++ b/docs/convert_maDLC.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- (convert-maDLC)= # How to convert a pre-2.2 project for use with DeepLabCut 2.2 or later diff --git a/docs/course.md b/docs/course.md index ff25b8d6e4..fa2eb30aba 100644 --- a/docs/course.md +++ b/docs/course.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- # DeepLabCut Self-paced Course ::::{warning} diff --git a/docs/dlc-live/deeplabcutlive.md b/docs/dlc-live/deeplabcutlive.md index 9b61f2f857..1d4b38d9ff 100644 --- a/docs/dlc-live/deeplabcutlive.md +++ b/docs/dlc-live/deeplabcutlive.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (deeplabcut-live)= # Running DeepLabCut models in real-time diff --git a/docs/dlc-live/dlc-live-gui/index.md b/docs/dlc-live/dlc-live-gui/index.md index ad5877061b..8f82588b2d 100644 --- a/docs/dlc-live/dlc-live-gui/index.md +++ b/docs/dlc-live/dlc-live-gui/index.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- # DeepLabCut-live-GUI A graphical application for **real-time pose estimation with DeepLabCut** using one or more cameras. @@ -71,5 +76,5 @@ Before getting started, be aware of the following constraints: ## Feedback, issues, and contributions > *This project is under active development. Feedback from real experimental use is highly valued.* -> +> > [Please report issues, suggest features, or contribute to the codebase on GitHub !](https://github.com/DeepLabCut/DeepLabCut-live-GUI) diff --git a/docs/dlc-live/dlc-live-gui/quickstart/install.md b/docs/dlc-live/dlc-live-gui/quickstart/install.md index 1fc163e7ca..fbbd385a28 100644 --- a/docs/dlc-live/dlc-live-gui/quickstart/install.md +++ b/docs/dlc-live/dlc-live-gui/quickstart/install.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- # Installation This page explains how to install **DeepLabCut-live-GUI** for interactive, real‑time pose estimation. diff --git a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/aravis_backend.md b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/aravis_backend.md index 18a061cf88..9cacbe8bb4 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/aravis_backend.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/aravis_backend.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (file:dlclivegui-camera-aravis-backend)= # Aravis backend diff --git a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/basler_backend.md b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/basler_backend.md index 0abc3dfd8e..7bd5b7097b 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/basler_backend.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/basler_backend.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (file:dlclivegui-basler-backend)= # Basler backend @@ -11,7 +16,7 @@ This backend requires the optional `pypylon` dependency. If `pypylon` is not ins --- -## Features & design +## Features & design - Native Basler camera support via **pypylon** (Pylon SDK bindings). - Best-effort device discovery without opening cameras (enumerates `DeviceInfo` entries). diff --git a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/camera_support.md b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/camera_support.md index a1f8b3b9a0..a3b7e1c440 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/camera_support.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/camera_support.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (file:dlclivegui-camera-support)= # Camera support @@ -32,9 +37,9 @@ You can select the backend in the GUI from the "Backend" dropdown, or in your co Below are some general recommendations for backend selection based on your operating system and camera type. ```{note} -Please understand this may not reflect the exact capabilities for every setup. +Please understand this may not reflect the exact capabilities for every setup. -Let us know about your experience with different cameras and backends on different platforms to help us improve our documentation and support! +Let us know about your experience with different cameras and backends on different platforms to help us improve our documentation and support! ``` ### Windows @@ -89,5 +94,3 @@ Install vendor-provided camera drivers and SDK. CTI files are typically in: | Windows | ✅ | ✅ | ❌ | ✅ | | Linux | ✅ | ✅ | ✅ | ✅ | | macOS | ✅ | ❌ | ⚠️ | ✅ | - - diff --git a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/gentl_backend.md b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/gentl_backend.md index cd94974bfd..84a53aed1a 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/gentl_backend.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/gentl_backend.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- # GenTL backend The GenTL backend provides support for **GenICam / GenTL** compatible cameras using the **Harvesters** Python library (a GenTL consumer). diff --git a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/opencv_backend.md b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/opencv_backend.md index 490b7a4ce5..df58205f5b 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/opencv_backend.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/cameras_backends/opencv_backend.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (file:dlclivegui-opencv-backend)= # OpenCV backend diff --git a/docs/dlc-live/dlc-live-gui/user_guide/misc/misc_landing.md b/docs/dlc-live/dlc-live-gui/user_guide/misc/misc_landing.md index 899cee8547..4bfab1712e 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/misc/misc_landing.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/misc/misc_landing.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- # Additional resources In this section, you can find additional resources related to the GUI and DLC-live, including: diff --git a/docs/dlc-live/dlc-live-gui/user_guide/misc/modelzoo_downloads.md b/docs/dlc-live/dlc-live-gui/user_guide/misc/modelzoo_downloads.md index 8ecba380dc..1cd04b34d9 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/misc/modelzoo_downloads.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/misc/modelzoo_downloads.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (file:dlclivegui-pretrained-models)= # Pre-trained models diff --git a/docs/dlc-live/dlc-live-gui/user_guide/misc/timestamp_format.md b/docs/dlc-live/dlc-live-gui/user_guide/misc/timestamp_format.md index 1b715fdeaa..b0b6958e1f 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/misc/timestamp_format.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/misc/timestamp_format.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- (file:dlclivegui-tinmestamp-format)= # Video timestamp format @@ -92,4 +97,4 @@ for frame_idx, timestamp in enumerate(data['timestamps']): The encoded video is written with a fixed input frame rate configured when recording starts. -The timestamps reflect capture/enqueue timing and may not perfectly match the encoded frame pacing, especially if frames are dropped or capture timing varies. \ No newline at end of file +The timestamps reflect capture/enqueue timing and may not perfectly match the encoded frame pacing, especially if frames are dropped or capture timing varies. diff --git a/docs/dlc-live/dlc-live-gui/user_guide/overview.md b/docs/dlc-live/dlc-live-gui/user_guide/overview.md index c9e5cc764d..fb975b5e22 100644 --- a/docs/dlc-live/dlc-live-gui/user_guide/overview.md +++ b/docs/dlc-live/dlc-live-gui/user_guide/overview.md @@ -1,3 +1,8 @@ +--- +deeplabcut: + last_metadata_updated: '2026-03-17' + ignore: false +--- # GUI overview DeepLabCut-live-GUI (`dlclivegui`) is a **PySide6-based desktop application** for running real-time DeepLabCut pose estimation experiments with **one or multiple cameras**, optional **processor plugins**, and **video recording** (with or without overlays). diff --git a/docs/docker.md b/docs/docker.md index 473cded14e..fc9048b86a 100644 --- a/docs/docker.md +++ b/docs/docker.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-04-15' + last_metadata_updated: '2026-03-06' + ignore: false +--- (docker-containers)= # DeepLabCut Docker containers diff --git a/docs/gui/PROJECT_GUI.md b/docs/gui/PROJECT_GUI.md index e0883c324e..362479a4b9 100644 --- a/docs/gui/PROJECT_GUI.md +++ b/docs/gui/PROJECT_GUI.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- (project-manager-gui)= # Interactive Project Manager GUI diff --git a/docs/gui/napari_GUI.md b/docs/gui/napari_GUI.md index df86f689bc..9acc9ddaea 100644 --- a/docs/gui/napari_GUI.md +++ b/docs/gui/napari_GUI.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- (napari-gui)= # napari labeling GUI diff --git a/docs/installation.md b/docs/installation.md index a9fb70051f..333a1e72bc 100644 --- a/docs/installation.md +++ b/docs/installation.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-23' + last_metadata_updated: '2026-03-06' + ignore: false +--- (how-to-install)= # How To Install DeepLabCut diff --git a/docs/intro.md b/docs/intro.md index 415e668dd5..cab6f4b183 100644 --- a/docs/intro.md +++ b/docs/intro.md @@ -1 +1,7 @@ +--- +deeplabcut: + last_content_updated: '2024-06-14' + last_metadata_updated: '2026-03-06' + ignore: false +--- Please see the main [READ ME!](https://deeplabcut.github.io/DeepLabCut/README.html) diff --git a/docs/maDLC_UserGuide.md b/docs/maDLC_UserGuide.md index 9861a41242..0de051dd92 100644 --- a/docs/maDLC_UserGuide.md +++ b/docs/maDLC_UserGuide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- (multi-animal-userguide)= # DeepLabCut for Multi-Animal Projects diff --git a/docs/pytorch/Benchmarking_shuffle_guide.md b/docs/pytorch/Benchmarking_shuffle_guide.md index 8e4554ce77..a498eea9bd 100644 --- a/docs/pytorch/Benchmarking_shuffle_guide.md +++ b/docs/pytorch/Benchmarking_shuffle_guide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- # DeepLabCut Benchmarking - User Guide ## Reasoning for benchmarking models in DLC (across DLC versions and architectures) diff --git a/docs/pytorch/architectures.md b/docs/pytorch/architectures.md index c1742b8221..aa4f8ed9f8 100644 --- a/docs/pytorch/architectures.md +++ b/docs/pytorch/architectures.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- (dlc3-architectures)= # DeepLabCut 3.0 - PyTorch Model Architectures diff --git a/docs/pytorch/pytorch_config.md b/docs/pytorch/pytorch_config.md index 8d75e3947e..a4ddf14599 100644 --- a/docs/pytorch/pytorch_config.md +++ b/docs/pytorch/pytorch_config.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-10-02' + last_metadata_updated: '2026-03-06' + ignore: false +--- (dlc3-pytorch-config)= # The PyTorch Configuration file diff --git a/docs/pytorch/user_guide.md b/docs/pytorch/user_guide.md index a30f7f6a7c..a9479d364b 100644 --- a/docs/pytorch/user_guide.md +++ b/docs/pytorch/user_guide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-07-01' + last_metadata_updated: '2026-03-06' + ignore: false +--- (dlc3-user-guide)= # DeepLabCut 3.0 - PyTorch User Guide diff --git a/docs/pytorch_dlc.md b/docs/pytorch_dlc.md index 157a1c19af..5da418aa87 100644 --- a/docs/pytorch_dlc.md +++ b/docs/pytorch_dlc.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2024-01-17' + last_metadata_updated: '2026-03-06' + ignore: false +--- # DeepLabCut: PyTorch API ## Modules diff --git a/docs/quick-start/single_animal_quick_guide.md b/docs/quick-start/single_animal_quick_guide.md index 307ec6d113..ef17102ab1 100644 --- a/docs/quick-start/single_animal_quick_guide.md +++ b/docs/quick-start/single_animal_quick_guide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- # QUICK GUIDE to single Animal Training: **The main steps to take you from project creation to analyzed videos:** diff --git a/docs/quick-start/tutorial_maDLC.md b/docs/quick-start/tutorial_maDLC.md index 507ad559d8..111ed996f5 100644 --- a/docs/quick-start/tutorial_maDLC.md +++ b/docs/quick-start/tutorial_maDLC.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Multi-animal pose estimation with DeepLabCut: A 5-minute tutorial ## GUI: diff --git a/docs/recipes/BatchProcessing.md b/docs/recipes/BatchProcessing.md index cd9dfef8a5..1e9a866e2a 100644 --- a/docs/recipes/BatchProcessing.md +++ b/docs/recipes/BatchProcessing.md @@ -1,4 +1,9 @@ - +--- +deeplabcut: + last_content_updated: '2025-09-16' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Automate training and video analysis: Batch Processing ## Tips for working with DLC networks: diff --git a/docs/recipes/ClusteringNapari.md b/docs/recipes/ClusteringNapari.md index a12fa75142..bb7faf41c9 100644 --- a/docs/recipes/ClusteringNapari.md +++ b/docs/recipes/ClusteringNapari.md @@ -1,4 +1,9 @@ - +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Clustering in the napari-DeepLabCut GUI To increase model performance, one can find the errors in the user-defined label (or in output H5 files after video diff --git a/docs/recipes/DLCMethods.md b/docs/recipes/DLCMethods.md index 74d5ec4c65..b3710b7d55 100644 --- a/docs/recipes/DLCMethods.md +++ b/docs/recipes/DLCMethods.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # How to write a DLC Methods Section **Pose estimation using DeepLabCut** diff --git a/docs/recipes/MegaDetectorDLCLive.md b/docs/recipes/MegaDetectorDLCLive.md index ecdf3432c7..e314b55600 100644 --- a/docs/recipes/MegaDetectorDLCLive.md +++ b/docs/recipes/MegaDetectorDLCLive.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- # 💚 MegaDetector+DeepLabCut 💜 [DeepLabCut-Live](https://github.com/DeepLabCut/DeepLabCut-live) is an open source and free real-time package from DeepLabCut that allows for real-time, low-latency pose estimation. [The DeepLabCut-ModelZoo](http://modelzoo.deeplabcut.org/) is our growing collection of pretrained animal models for rapid deployment; no training is typically required to use these models. MegaDetector is a free open software trained to detect animals, people, and vehicles from camera trap images. Check [here](https://github.com/microsoft/CameraTraps/blob/main/megadetector.md) for further information. diff --git a/docs/recipes/OpenVINO.md b/docs/recipes/OpenVINO.md index 78ea18e82f..06bcbba3b5 100644 --- a/docs/recipes/OpenVINO.md +++ b/docs/recipes/OpenVINO.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Intel OpenVINO backend ::::{warning} diff --git a/docs/recipes/OtherData.md b/docs/recipes/OtherData.md index 73343284b3..ebc9b16f3e 100644 --- a/docs/recipes/OtherData.md +++ b/docs/recipes/OtherData.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # How to use data labeled outside of DeepLabCut - and/or if you merge projects across scorers (see below): diff --git a/docs/recipes/TechHardware.md b/docs/recipes/TechHardware.md index 6fb1add9bc..c32a6d741d 100644 --- a/docs/recipes/TechHardware.md +++ b/docs/recipes/TechHardware.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2026-02-10' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Technical (Hardware) Considerations ## Quick summary: diff --git a/docs/recipes/UsingModelZooPupil.md b/docs/recipes/UsingModelZooPupil.md index d73a1acbd7..2b906ad755 100644 --- a/docs/recipes/UsingModelZooPupil.md +++ b/docs/recipes/UsingModelZooPupil.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Using ModelZoo models on your own datasets

Animal behavior has to be analyzed with painstaking accuracy. Therefore, animal pose estimation has been diff --git a/docs/recipes/flip_and_rotate.ipynb b/docs/recipes/flip_and_rotate.ipynb index 501b7969d9..55deba2fe7 100644 --- a/docs/recipes/flip_and_rotate.ipynb +++ b/docs/recipes/flip_and_rotate.ipynb @@ -1867,6 +1867,11 @@ ], "metadata": { "celltoolbar": "Edit Metadata", + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-16", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python [conda env:DEEPLABCUT_newGUI] *", "language": "python", diff --git a/docs/recipes/installTips.md b/docs/recipes/installTips.md index 28ee27a56e..1ddefa0ca7 100644 --- a/docs/recipes/installTips.md +++ b/docs/recipes/installTips.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- (installation-tips)= # Installation Tips diff --git a/docs/recipes/io.md b/docs/recipes/io.md index e97238628b..f18c4d28b6 100644 --- a/docs/recipes/io.md +++ b/docs/recipes/io.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2022-04-11' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Input/output manipulations with DeepLabCut ## Analyzing very large videos in chunks diff --git a/docs/recipes/nn.md b/docs/recipes/nn.md index 9377c446ca..3ca09cf14b 100644 --- a/docs/recipes/nn.md +++ b/docs/recipes/nn.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- (tf-training-tips-and-tricks)= # Model training tips & tricks diff --git a/docs/recipes/pose_cfg_file_breakdown.md b/docs/recipes/pose_cfg_file_breakdown.md index 2f79ac28d3..dbe97ea273 100644 --- a/docs/recipes/pose_cfg_file_breakdown.md +++ b/docs/recipes/pose_cfg_file_breakdown.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- # The `pose_cfg.yaml` Guideline Handbook ::::{warning} diff --git a/docs/recipes/post.md b/docs/recipes/post.md index cbcb78c673..1ebdf09ec5 100644 --- a/docs/recipes/post.md +++ b/docs/recipes/post.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2022-06-08' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Some data processing recipes! ## Flagging frames with abnormal bodypart distances diff --git a/docs/recipes/publishing_notebooks_into_the_DLC_main_cookbook.md b/docs/recipes/publishing_notebooks_into_the_DLC_main_cookbook.md index 83dbb8c75e..ca2ff2c727 100644 --- a/docs/recipes/publishing_notebooks_into_the_DLC_main_cookbook.md +++ b/docs/recipes/publishing_notebooks_into_the_DLC_main_cookbook.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- # Publishing Notebooks into the Main DLC Cookbook ### Your Recipe Guide to Contributing to the DLC Cookbook diff --git a/docs/roadmap.md b/docs/roadmap.md index b303754e37..04228cb56e 100644 --- a/docs/roadmap.md +++ b/docs/roadmap.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-02-28' + last_metadata_updated: '2026-03-06' + ignore: false +--- (dev-roadmap)= ## A development roadmap for DeepLabCut diff --git a/docs/standardDeepLabCut_UserGuide.md b/docs/standardDeepLabCut_UserGuide.md index f7e653e487..09812b7050 100644 --- a/docs/standardDeepLabCut_UserGuide.md +++ b/docs/standardDeepLabCut_UserGuide.md @@ -1,3 +1,9 @@ +--- +deeplabcut: + last_content_updated: '2025-06-30' + last_metadata_updated: '2026-03-06' + ignore: false +--- (single-animal-userguide)= # DeepLabCut User Guide (for single animal projects) diff --git a/examples/COLAB/COLAB_3miceDemo.ipynb b/examples/COLAB/COLAB_3miceDemo.ipynb index 427602ff4c..43a6260a07 100644 --- a/examples/COLAB/COLAB_3miceDemo.ipynb +++ b/examples/COLAB/COLAB_3miceDemo.ipynb @@ -256,6 +256,11 @@ "name": "Copy of 3micedemo.ipynb", "provenance": [] }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2026-02-10", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3", "name": "python3" diff --git a/examples/COLAB/COLAB_BUCTD_and_CTD_tracking.ipynb b/examples/COLAB/COLAB_BUCTD_and_CTD_tracking.ipynb index e7e123ca23..43b7def161 100644 --- a/examples/COLAB/COLAB_BUCTD_and_CTD_tracking.ipynb +++ b/examples/COLAB/COLAB_BUCTD_and_CTD_tracking.ipynb @@ -2358,6 +2358,11 @@ "gpuType": "T4", "provenance": [] }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-10-02", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3", "name": "python3" diff --git a/examples/COLAB/COLAB_DEMO_SuperAnimal.ipynb b/examples/COLAB/COLAB_DEMO_SuperAnimal.ipynb index f6efebe8b0..c0f5c70f46 100644 --- a/examples/COLAB/COLAB_DEMO_SuperAnimal.ipynb +++ b/examples/COLAB/COLAB_DEMO_SuperAnimal.ipynb @@ -213,6 +213,11 @@ "colab": { "provenance": [] }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-06-30", + "last_metadata_updated": "2026-03-06" + }, "gpuClass": "standard", "kernelspec": { "display_name": "dlc", diff --git a/examples/COLAB/COLAB_DEMO_mouse_openfield.ipynb b/examples/COLAB/COLAB_DEMO_mouse_openfield.ipynb index a4d7543147..139e723c0f 100644 --- a/examples/COLAB/COLAB_DEMO_mouse_openfield.ipynb +++ b/examples/COLAB/COLAB_DEMO_mouse_openfield.ipynb @@ -310,6 +310,11 @@ "name": "Colab_DEMO_mouse_openfield.ipynb", "provenance": [] }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-16", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python [default]", "language": "python", diff --git a/examples/COLAB/COLAB_DLC_ModelZoo.ipynb b/examples/COLAB/COLAB_DLC_ModelZoo.ipynb index 1c312afd22..79b655b915 100644 --- a/examples/COLAB/COLAB_DLC_ModelZoo.ipynb +++ b/examples/COLAB/COLAB_DLC_ModelZoo.ipynb @@ -1,315 +1,320 @@ { - "cells": [ - { - "cell_type": "markdown", - "metadata": { - "colab_type": "text", - "id": "view-in-github" - }, - "source": [ - "\"Open" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "RK255E7YoEIt" - }, - "source": [ - "# DeepLabCut Model Zoo user-contributed models\n", - "\n", - "🚨 **WARNING** -- This is using the old version from 2020-2023 with user-supplied models. Please see the SuperAnimal notebook if you want to use our Foundational Models for Quadrupeds or mice.\n", - "\n", - "![alt text](https://images.squarespace-cdn.com/content/v1/57f6d51c9f74566f55ecf271/1616492373700-PGOAC72IOB6AUE47VTJX/ke17ZwdGBToddI8pDm48kB8JrdUaZR-OSkKLqWQPp_YUqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8N_N4V1vUb5AoIIIbLZhVYwL8IeDg6_3B-BRuF4nNrNcQkVuAT7tdErd0wQFEGFSnBqyW03PFN2MN6T6ry5cmXqqA9xITfsbVGDrg_goIDasRCalqV8R3606BuxERAtDaQ/modelzoo.png?format=1000w)\n", - "\n", - "http://modelzoo.deeplabcut.org\n", - "\n", - "You can use this notebook to analyze videos with pretrained networks from our model zoo - NO local installation of DeepLabCut is needed!\n", - "\n", - "- **What you need:** a video of your favorite dog, cat, human, etc: check the list of currently available models here: http://modelzoo.deeplabcut.org\n", - "\n", - "- **What to do:** (1) in the top right corner, click \"CONNECT\". Then, just hit run (play icon) on each cell below and follow the instructions!\n", - "\n", - "## **Please consider giving back and labeling a little data to help make each network even better!**\n", - "\n", - "We have a WebApp, so no need to install anything, just a few clicks! We'd really appreciate your help!\n", - " \n", - "https://contrib.deeplabcut.org/\n", - "\n", - "\n", - "- **Note, if you performance is less that you would like:** firstly check the labeled_video parameters (i.e. \"pcutoff\" in the config.yaml file that will set the video plotting) - see the end of this notebook. You can also use the model in your own projects locally. Please be sure to cite the papers for the model, and http://modelzoo.deeplabcut.org (paper forthcoming!)\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "## **Let's get going: install DeepLabCut into COLAB:**\n", - "\n", - "*Also, be sure you are connected to a GPU: go to menu, click Runtime > Change Runtime Type > select \"GPU\"*\n" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": {}, - "outputs": [], - "source": [ - "# Install the latest version of DeepLabCut\n", - "!pip install --pre \"deeplabcut[tf,modelzoo]\"" - ] - }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### Important - Restart the Runtime for the updated packages to be imported!\n", - "\n", - "PLEASE, click \"restart runtime\" from the output above before proceeding!" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "ZT4PwGSbYQEO" - }, - "source": [ - "## Now let's set the backend & import the DeepLabCut package\n", - "### (if colab is buggy/throws an error, just rerun this cell):" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "bvoiWefrYQEP" - }, - "outputs": [], - "source": [ - "import os\n", - "import deeplabcut" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "syweXs88tyuO" - }, - "source": [ - "## Next, run the cell below to upload your video file from your computer:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "7eqEZYs_CaLy" - }, - "outputs": [], - "source": [ - "from google.colab import files\n", - "\n", - "uploaded = files.upload()\n", - "for filepath, content in uploaded.items():\n", - " print(f'User uploaded file \"{filepath}\" with length {len(content)} bytes')\n", - "video_path = os.path.abspath(filepath)\n", - "\n", - "# If this cell fails (e.g., when using Safari in place of Google Chrome),\n", - "# manually upload your video via the Files menu to the left\n", - "# and define `video_path` yourself with right click > copy path on the video." - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "YsaqOTkZtf-w" - }, - "source": [ - "## Select your model from the dropdown menu, then below (optionally) input the name you want for the project:\n" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "Ih0t7lUjYQEd" - }, - "outputs": [], - "source": [ - "import ipywidgets as widgets\n", - "from IPython.display import display\n", - "\n", - "model_options = deeplabcut.create_project.modelzoo.Modeloptions\n", - "model_selection = widgets.Dropdown(\n", - " options=model_options,\n", - " value=model_options[0],\n", - " description=\"Choose a DLC ModelZoo model!\",\n", - " disabled=False\n", - ")\n", - "display(model_selection)" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "UV0QXswGCFrI" - }, - "outputs": [], - "source": [ - "project_name = 'myDLC_modelZoo'\n", - "your_name = 'teamDLC'\n", - "model2use = model_selection.value\n", - "videotype = os.path.splitext(video_path)[-1].lstrip('.') #or MOV, or avi, whatever you uploaded!" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "JQxko-t3uMVO" - }, - "source": [ - "## Attention on this step !!\n", - "- Please note that for optimal performance your videos should contain frames that are around ~300-600 pixels (on one edge). If you have a larger video (like from an iPhone, first downsize by running this please! :)\n", - "\n", - "- Thus, if you're using an iPhone, or such, you'll need to downsample the video first by running the code below**\n", - "\n", - "(no need to edit it unless you want to change the size)" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "WpAX3BKY94e0" - }, - "outputs": [], - "source": [ - "video_path = deeplabcut.DownSampleVideo(video_path, width=300)\n", - "print(video_path)" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "KJm_Vbx-s5OY" - }, - "source": [ - "## Lastly, run the cell below to create a pretrained project, analyze your video with your selected pretrained network, plot trajectories, and create a labeled video!:\n" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "T9MGgAdIFKPY" - }, - "outputs": [], - "source": [ - "config_path, train_config_path = deeplabcut.create_pretrained_project(\n", - " project_name,\n", - " your_name,\n", - " [video_path],\n", - " videotype=videotype,\n", - " model=model2use,\n", - " analyzevideo=True,\n", - " createlabeledvideo=True,\n", - " copy_videos=True, #must leave copy_videos=True\n", - " engine=deeplabcut.Engine.TF,\n", - ")" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "WS-KxhBMvEBj" - }, - "source": [ - "Now, you can move this project from Colab (i.e. download it to your GoogleDrive), and use it like a normal standard project!\n", - "\n", - "You can analyze more videos, extract outliers, refine then, and/or then add new key points + label new frames, and retrain if desired. We hope this gives you a good launching point for your work!\n", - "\n", - "###Happy DeepLabCutting! Welcome to the Zoo :)\n", - "\n" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "KPOqiLmo6d7t" - }, - "source": [ - "## More advanced options:\n", - "\n", - "- If you would now like to customize the video/plots - i.e., color, dot size, threshold for the point to be plotted (pcutoff), please simply edit the \"config.yaml\" file by updating the values below:" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "yGLNVK1q6rIp" - }, - "outputs": [], - "source": [ - "# Updating the plotting within the config.yaml file (without opening it ;):\n", - "edits = {\n", - " 'dotsize': 7, # size of the dots!\n", - " 'colormap': 'spring', # any matplotlib colormap!\n", - " 'pcutoff': 0.5, # the higher the more conservative the plotting!\n", - "}\n", - "deeplabcut.auxiliaryfunctions.edit_config(config_path, edits)" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "Vlc0wZgB7R5e" - }, - "outputs": [], - "source": [ - "# re-create the labeled video (first you will need to delete in the folder to the LEFT!):\n", - "project_path = os.path.dirname(config_path)\n", - "full_video_path = os.path.join(\n", - " project_path,\n", - " 'videos',\n", - " os.path.basename(video_path),\n", - ")\n", - "\n", - "#filter predictions (should already be done above ;):\n", - "deeplabcut.filterpredictions(config_path, [full_video_path], videotype=videotype)\n", - "\n", - "#re-create the video with your edits!\n", - "deeplabcut.create_labeled_video(config_path, [full_video_path], videotype=videotype, filtered=True)" - ] - } - ], - "metadata": { - "colab": { - "include_colab_link": true, - "name": "Copy of COLAB_DLC_ModelZoo.ipynb", - "provenance": [], - "toc_visible": true - }, - "gpuClass": "standard", - "kernelspec": { - "display_name": "Python 3", - "language": "python", - "name": "python3" - }, - "language_info": { - "codemirror_mode": { - "name": "ipython", - "version": 3 - }, - "file_extension": ".py", - "mimetype": "text/x-python", - "name": "python", - "nbconvert_exporter": "python", - "pygments_lexer": "ipython3", - "version": "3.7.7" - } + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "colab_type": "text", + "id": "view-in-github" + }, + "source": [ + "\"Open" + ] }, - "nbformat": 4, - "nbformat_minor": 0 + { + "cell_type": "markdown", + "metadata": { + "id": "RK255E7YoEIt" + }, + "source": [ + "# DeepLabCut Model Zoo user-contributed models\n", + "\n", + "🚨 **WARNING** -- This is using the old version from 2020-2023 with user-supplied models. Please see the SuperAnimal notebook if you want to use our Foundational Models for Quadrupeds or mice.\n", + "\n", + "![alt text](https://images.squarespace-cdn.com/content/v1/57f6d51c9f74566f55ecf271/1616492373700-PGOAC72IOB6AUE47VTJX/ke17ZwdGBToddI8pDm48kB8JrdUaZR-OSkKLqWQPp_YUqsxRUqqbr1mOJYKfIPR7LoDQ9mXPOjoJoqy81S2I8N_N4V1vUb5AoIIIbLZhVYwL8IeDg6_3B-BRuF4nNrNcQkVuAT7tdErd0wQFEGFSnBqyW03PFN2MN6T6ry5cmXqqA9xITfsbVGDrg_goIDasRCalqV8R3606BuxERAtDaQ/modelzoo.png?format=1000w)\n", + "\n", + "http://modelzoo.deeplabcut.org\n", + "\n", + "You can use this notebook to analyze videos with pretrained networks from our model zoo - NO local installation of DeepLabCut is needed!\n", + "\n", + "- **What you need:** a video of your favorite dog, cat, human, etc: check the list of currently available models here: http://modelzoo.deeplabcut.org\n", + "\n", + "- **What to do:** (1) in the top right corner, click \"CONNECT\". Then, just hit run (play icon) on each cell below and follow the instructions!\n", + "\n", + "## **Please consider giving back and labeling a little data to help make each network even better!**\n", + "\n", + "We have a WebApp, so no need to install anything, just a few clicks! We'd really appreciate your help!\n", + " \n", + "https://contrib.deeplabcut.org/\n", + "\n", + "\n", + "- **Note, if you performance is less that you would like:** firstly check the labeled_video parameters (i.e. \"pcutoff\" in the config.yaml file that will set the video plotting) - see the end of this notebook. You can also use the model in your own projects locally. Please be sure to cite the papers for the model, and http://modelzoo.deeplabcut.org (paper forthcoming!)\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "## **Let's get going: install DeepLabCut into COLAB:**\n", + "\n", + "*Also, be sure you are connected to a GPU: go to menu, click Runtime > Change Runtime Type > select \"GPU\"*\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "# Install the latest version of DeepLabCut\n", + "!pip install --pre \"deeplabcut[tf,modelzoo]\"" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Important - Restart the Runtime for the updated packages to be imported!\n", + "\n", + "PLEASE, click \"restart runtime\" from the output above before proceeding!" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "ZT4PwGSbYQEO" + }, + "source": [ + "## Now let's set the backend & import the DeepLabCut package\n", + "### (if colab is buggy/throws an error, just rerun this cell):" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "bvoiWefrYQEP" + }, + "outputs": [], + "source": [ + "import os\n", + "import deeplabcut" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "syweXs88tyuO" + }, + "source": [ + "## Next, run the cell below to upload your video file from your computer:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "7eqEZYs_CaLy" + }, + "outputs": [], + "source": [ + "from google.colab import files\n", + "\n", + "uploaded = files.upload()\n", + "for filepath, content in uploaded.items():\n", + " print(f'User uploaded file \"{filepath}\" with length {len(content)} bytes')\n", + "video_path = os.path.abspath(filepath)\n", + "\n", + "# If this cell fails (e.g., when using Safari in place of Google Chrome),\n", + "# manually upload your video via the Files menu to the left\n", + "# and define `video_path` yourself with right click > copy path on the video." + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "YsaqOTkZtf-w" + }, + "source": [ + "## Select your model from the dropdown menu, then below (optionally) input the name you want for the project:\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "Ih0t7lUjYQEd" + }, + "outputs": [], + "source": [ + "import ipywidgets as widgets\n", + "from IPython.display import display\n", + "\n", + "model_options = deeplabcut.create_project.modelzoo.Modeloptions\n", + "model_selection = widgets.Dropdown(\n", + " options=model_options,\n", + " value=model_options[0],\n", + " description=\"Choose a DLC ModelZoo model!\",\n", + " disabled=False\n", + ")\n", + "display(model_selection)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "UV0QXswGCFrI" + }, + "outputs": [], + "source": [ + "project_name = 'myDLC_modelZoo'\n", + "your_name = 'teamDLC'\n", + "model2use = model_selection.value\n", + "videotype = os.path.splitext(video_path)[-1].lstrip('.') #or MOV, or avi, whatever you uploaded!" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "JQxko-t3uMVO" + }, + "source": [ + "## Attention on this step !!\n", + "- Please note that for optimal performance your videos should contain frames that are around ~300-600 pixels (on one edge). If you have a larger video (like from an iPhone, first downsize by running this please! :)\n", + "\n", + "- Thus, if you're using an iPhone, or such, you'll need to downsample the video first by running the code below**\n", + "\n", + "(no need to edit it unless you want to change the size)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "WpAX3BKY94e0" + }, + "outputs": [], + "source": [ + "video_path = deeplabcut.DownSampleVideo(video_path, width=300)\n", + "print(video_path)" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "KJm_Vbx-s5OY" + }, + "source": [ + "## Lastly, run the cell below to create a pretrained project, analyze your video with your selected pretrained network, plot trajectories, and create a labeled video!:\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "T9MGgAdIFKPY" + }, + "outputs": [], + "source": [ + "config_path, train_config_path = deeplabcut.create_pretrained_project(\n", + " project_name,\n", + " your_name,\n", + " [video_path],\n", + " videotype=videotype,\n", + " model=model2use,\n", + " analyzevideo=True,\n", + " createlabeledvideo=True,\n", + " copy_videos=True, #must leave copy_videos=True\n", + " engine=deeplabcut.Engine.TF,\n", + ")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "WS-KxhBMvEBj" + }, + "source": [ + "Now, you can move this project from Colab (i.e. download it to your GoogleDrive), and use it like a normal standard project!\n", + "\n", + "You can analyze more videos, extract outliers, refine then, and/or then add new key points + label new frames, and retrain if desired. We hope this gives you a good launching point for your work!\n", + "\n", + "###Happy DeepLabCutting! Welcome to the Zoo :)\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "KPOqiLmo6d7t" + }, + "source": [ + "## More advanced options:\n", + "\n", + "- If you would now like to customize the video/plots - i.e., color, dot size, threshold for the point to be plotted (pcutoff), please simply edit the \"config.yaml\" file by updating the values below:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "yGLNVK1q6rIp" + }, + "outputs": [], + "source": [ + "# Updating the plotting within the config.yaml file (without opening it ;):\n", + "edits = {\n", + " 'dotsize': 7, # size of the dots!\n", + " 'colormap': 'spring', # any matplotlib colormap!\n", + " 'pcutoff': 0.5, # the higher the more conservative the plotting!\n", + "}\n", + "deeplabcut.auxiliaryfunctions.edit_config(config_path, edits)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "Vlc0wZgB7R5e" + }, + "outputs": [], + "source": [ + "# re-create the labeled video (first you will need to delete in the folder to the LEFT!):\n", + "project_path = os.path.dirname(config_path)\n", + "full_video_path = os.path.join(\n", + " project_path,\n", + " 'videos',\n", + " os.path.basename(video_path),\n", + ")\n", + "\n", + "#filter predictions (should already be done above ;):\n", + "deeplabcut.filterpredictions(config_path, [full_video_path], videotype=videotype)\n", + "\n", + "#re-create the video with your edits!\n", + "deeplabcut.create_labeled_video(config_path, [full_video_path], videotype=videotype, filtered=True)" + ] + } + ], + "metadata": { + "colab": { + "include_colab_link": true, + "name": "Copy of COLAB_DLC_ModelZoo.ipynb", + "provenance": [], + "toc_visible": true + }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-10-02", + "last_metadata_updated": "2026-03-06" + }, + "gpuClass": "standard", + "kernelspec": { + "display_name": "Python 3", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.7.7" + } + }, + "nbformat": 4, + "nbformat_minor": 0 } diff --git a/examples/COLAB/COLAB_HumanPose_with_RTMPose.ipynb b/examples/COLAB/COLAB_HumanPose_with_RTMPose.ipynb index f3a13ac588..6d3f4be32d 100644 --- a/examples/COLAB/COLAB_HumanPose_with_RTMPose.ipynb +++ b/examples/COLAB/COLAB_HumanPose_with_RTMPose.ipynb @@ -1,1182 +1,1187 @@ { - "cells": [ - { - "cell_type": "markdown", - "metadata": { - "id": "t3P1R5BTwud1" - }, - "source": [ - "\"Open\n", - "\n", - "# DeepLabCut RTMPose human pose estimation demo" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "tJm8QpTzyAEe" - }, - "source": [ - "Some useful links:\n", - "\n", - "- DeepLabCut's GitHub: [github.com/DeepLabCut/DeepLabCut](https://github.com/DeepLabCut/DeepLabCut/tree/main)\n", - "- DeepLabCut's Documentation: [deeplabcut.github.io/DeepLabCut](https://deeplabcut.github.io/DeepLabCut/README.html)\n", - "\n", - "This notebook illustrates how to use the cloud to run pose estimation on humans using a pre-trained [RTMPose](https://arxiv.org/abs/2303.07399) model. **⚠️Note: It uses DeepLabCut's low-level interface, so may be suited for more experienced users.⚠️**\n", - "\n", - "RTMPose is a top-down pose estimation model, which means that bounding boxes must be obtained for individuals (which is usually done through an [object detection model](https://en.wikipedia.org/wiki/Object_detection)) before running pose estimation. We obtain bounding boxes using a pre-trained object detector provided by [`torchvision`](https://pytorch.org/vision/main/models.html#object-detection-instance-segmentation-and-person-keypoint-detection).\n", - "\n", - "## Selecting the Runtime and Installing DeepLabCut\n", - "\n", - "**First, go to \"Runtime\" ->\"change runtime type\"->select \"Python3\", and then select \"GPU\".**\n", - "\n", - "Next, we need to install DeepLabCut and its dependencies." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "Aj7Fgm0Xx_fS" - }, - "outputs": [], - "source": [ - "# this will take a couple of minutes to install all the dependencies!\n", - "!pip install --pre deeplabcut" - ] - }, - { - "cell_type": "markdown", - "metadata": { - "id": "twiCWHbgzbwH" - }, - "source": [ - "**(Be sure to click \"RESTART RUNTIME\" if it is displayed above before moving on !) You will see this button at the output of the cells above ^.**" - ] + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "t3P1R5BTwud1" + }, + "source": [ + "\"Open\n", + "\n", + "# DeepLabCut RTMPose human pose estimation demo" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "tJm8QpTzyAEe" + }, + "source": [ + "Some useful links:\n", + "\n", + "- DeepLabCut's GitHub: [github.com/DeepLabCut/DeepLabCut](https://github.com/DeepLabCut/DeepLabCut/tree/main)\n", + "- DeepLabCut's Documentation: [deeplabcut.github.io/DeepLabCut](https://deeplabcut.github.io/DeepLabCut/README.html)\n", + "\n", + "This notebook illustrates how to use the cloud to run pose estimation on humans using a pre-trained [RTMPose](https://arxiv.org/abs/2303.07399) model. **⚠️Note: It uses DeepLabCut's low-level interface, so may be suited for more experienced users.⚠️**\n", + "\n", + "RTMPose is a top-down pose estimation model, which means that bounding boxes must be obtained for individuals (which is usually done through an [object detection model](https://en.wikipedia.org/wiki/Object_detection)) before running pose estimation. We obtain bounding boxes using a pre-trained object detector provided by [`torchvision`](https://pytorch.org/vision/main/models.html#object-detection-instance-segmentation-and-person-keypoint-detection).\n", + "\n", + "## Selecting the Runtime and Installing DeepLabCut\n", + "\n", + "**First, go to \"Runtime\" ->\"change runtime type\"->select \"Python3\", and then select \"GPU\".**\n", + "\n", + "Next, we need to install DeepLabCut and its dependencies." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "Aj7Fgm0Xx_fS" + }, + "outputs": [], + "source": [ + "# this will take a couple of minutes to install all the dependencies!\n", + "!pip install --pre deeplabcut" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "twiCWHbgzbwH" + }, + "source": [ + "**(Be sure to click \"RESTART RUNTIME\" if it is displayed above before moving on !) You will see this button at the output of the cells above ^.**" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "x6DugzWMzGoj" + }, + "source": [ + "## Importing Packages and Downloading Model Snapshots" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Y7jKbk_mzPJR" + }, + "source": [ + "Next, we'll need to import `deeplabcut`, `huggingface_hub` and other dependencies needed to run the demo." + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "gbXwpGKXzF98", + "outputId": "d7cc8390-e76a-4cc6-b945-42f0951c8d01" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "id": "x6DugzWMzGoj" - }, - "source": [ - "## Importing Packages and Downloading Model Snapshots" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Loading DLC 3.0.0rc10...\n", + "DLC loaded in light mode; you cannot use any GUI (labeling, relabeling and standalone GUI)\n" + ] + } + ], + "source": [ + "from pathlib import Path\n", + "\n", + "import deeplabcut.pose_estimation_pytorch as dlc_torch\n", + "import huggingface_hub\n", + "import matplotlib.collections as collections\n", + "import matplotlib.pyplot as plt\n", + "import numpy as np\n", + "import torch\n", + "import torchvision.models.detection as detection\n", + "from PIL import Image\n", + "from tqdm import tqdm" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "6KWKmWRxzX5R" + }, + "source": [ + "We can now download the pre-trained RTMPose model weights with which we'll run pose estimation." + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "L_V11iCszw3s", + "outputId": "8b010e6c-27f5-46ad-f713-2fd07effa3b1" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "id": "Y7jKbk_mzPJR" - }, - "source": [ - "Next, we'll need to import `deeplabcut`, `huggingface_hub` and other dependencies needed to run the demo." - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/huggingface_hub/utils/_auth.py:94: UserWarning: \n", + "The secret `HF_TOKEN` does not exist in your Colab secrets.\n", + "To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.\n", + "You will be able to reuse this secret in all of your notebooks.\n", + "Please note that authentication is recommended but still optional to access public models or datasets.\n", + " warnings.warn(\n" + ] + } + ], + "source": [ + "# Folder in COLAB where snapshots will be saved\n", + "model_files = Path(\"hf_files\").resolve()\n", + "model_files.mkdir(exist_ok=True)\n", + "\n", + "# Download the snapshot and model configuration file\n", + "# This is generic code to download any snapshot from HuggingFace\n", + "# To download DeepLabCut SuperAnimal or Model Zoo models, check\n", + "# out dlclibrary!\n", + "path_model_config = Path(\n", + " huggingface_hub.hf_hub_download(\n", + " \"DeepLabCut/HumanBody\",\n", + " \"rtmpose-x_simcc-body7_pytorch_config.yaml\",\n", + " local_dir=model_files,\n", + " )\n", + ")\n", + "path_snapshot = Path(\n", + " huggingface_hub.hf_hub_download(\n", + " \"DeepLabCut/HumanBody\",\n", + " \"rtmpose-x_simcc-body7.pt\",\n", + " local_dir=model_files,\n", + " )\n", + ")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "eEqukXXy0coy" + }, + "source": [ + "We'll now also define some parameters that we'll later use to plot predictions:\n", + "\n", + "- a colormap for the keypoints to plot\n", + "- a colormap for the limbs of the skeleton\n", + "- a skeleton for the model\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": { + "id": "Tam4rfJK0c_b" + }, + "outputs": [], + "source": [ + "cmap_keypoints = plt.get_cmap(\"rainbow\")\n", + "cmap_skeleton = plt.get_cmap(\"rainbow_r\")\n", + "\n", + "bodyparts2connect = [\n", + " (\"right_ankle\", \"right_knee\"),\n", + " (\"right_knee\", \"right_hip\"),\n", + " (\"left_ankle\", \"left_knee\"),\n", + " (\"left_hip\", \"left_knee\"),\n", + " (\"left_hip\", \"right_hip\"),\n", + " (\"right_shoulder\", \"right_hip\"),\n", + " (\"left_shoulder\", \"left_hip\"),\n", + " (\"left_shoulder\", \"right_shoulder\"),\n", + " (\"left_shoulder\", \"left_elbow\"),\n", + " (\"right_shoulder\", \"right_elbow\"),\n", + " (\"left_elbow\", \"left_wrist\"),\n", + " (\"right_elbow\", \"right_wrist\"),\n", + " (\"right_eye\", \"left_ear\"),\n", + " (\"left_eye\", \"right_eye\"),\n", + " (\"left_eye\", \"left_ear\"),\n", + " (\"right_eye\", \"right_ear\"),\n", + " (\"left_ear\", \"left_shoulder\"),\n", + " (\"right_ear\", \"right_shoulder\"),\n", + " (\"left_shoulder\", \"left_elbow\"),\n", + " (\"right_shoulder\", \"right_elbow\"),\n", + "]\n", + "skeleton = [\n", + " [16, 14],\n", + " [14, 12],\n", + " [17, 15],\n", + " [15, 13],\n", + " [12, 13],\n", + " [6, 12],\n", + " [7, 13],\n", + " [6, 7],\n", + " [6, 8],\n", + " [7, 9],\n", + " [8, 10],\n", + " [9, 11],\n", + " [2, 3],\n", + " [1, 2],\n", + " [1, 3],\n", + " [2, 4],\n", + " [3, 5],\n", + " [4, 6],\n", + " [5, 7],\n", + "]" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "cCxkkd-b0EJq" + }, + "source": [ + "## Running Inference on Images" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "dotn_xN-05gh" + }, + "source": [ + "First, let's upload some images to run inference on. To do so, you can just run the cell below." + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 92 }, + "id": "mZtikE1H0D34", + "outputId": "3d47314f-3ed0-40b2-e54d-2677feef9943" + }, + "outputs": [ { - "cell_type": "code", - "execution_count": 2, - "metadata": { - "id": "gbXwpGKXzF98", - "outputId": "d7cc8390-e76a-4cc6-b945-42f0951c8d01", - "colab": { - "base_uri": "https://localhost:8080/" - } - }, - "outputs": [ - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Loading DLC 3.0.0rc10...\n", - "DLC loaded in light mode; you cannot use any GUI (labeling, relabeling and standalone GUI)\n" - ] - } + "data": { + "text/html": [ + "\n", + " \n", + " \n", + " Upload widget is only available when the cell has been executed in the\n", + " current browser session. Please rerun this cell to enable.\n", + " \n", + " " ], - "source": [ - "from pathlib import Path\n", - "\n", - "import deeplabcut.pose_estimation_pytorch as dlc_torch\n", - "import huggingface_hub\n", - "import matplotlib.collections as collections\n", - "import matplotlib.pyplot as plt\n", - "import numpy as np\n", - "import torch\n", - "import torchvision.models.detection as detection\n", - "from PIL import Image\n", - "from tqdm import tqdm" + "text/plain": [ + "" ] + }, + "metadata": {}, + "output_type": "display_data" }, { - "cell_type": "markdown", - "metadata": { - "id": "6KWKmWRxzX5R" - }, - "source": [ - "We can now download the pre-trained RTMPose model weights with which we'll run pose estimation." - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Saving taylor_swift.jpg to taylor_swift.jpg\n", + "User uploaded file 'taylor_swift.jpg' with length 46915 bytes\n" + ] + } + ], + "source": [ + "from google.colab import files\n", + "\n", + "#JPG or PNG is recommended:\n", + "uploaded = files.upload()\n", + "for filepath, content in uploaded.items():\n", + " print(f\"User uploaded file '{filepath}' with length {len(content)} bytes\")\n", + "\n", + "image_paths = [Path(filepath).resolve() for filepath in uploaded.keys()]\n", + "\n", + "# If this cell fails (e.g., when using Safari in place of Google Chrome),\n", + "# manually upload your image via the Files menu to the left and define\n", + "# `image_paths` yourself with right `click` > `copy path` on the image:\n", + "#\n", + "# image_paths = [\n", + "# Path(\"/path/to/my/image_000.png\"),\n", + "# Path(\"/path/to/my/image_001.png\"),\n", + "# ]\n" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "nj-HtOBSwtdk", + "outputId": "eb5f3b18-cc89-4dd1-a58e-6c39c62582af" + }, + "outputs": [ { - "cell_type": "code", - "execution_count": 3, - "metadata": { - "id": "L_V11iCszw3s", - "outputId": "8b010e6c-27f5-46ad-f713-2fd07effa3b1", - "colab": { - "base_uri": "https://localhost:8080/" - } - }, - "outputs": [ - { - "output_type": "stream", - "name": "stderr", - "text": [ - "/usr/local/lib/python3.11/dist-packages/huggingface_hub/utils/_auth.py:94: UserWarning: \n", - "The secret `HF_TOKEN` does not exist in your Colab secrets.\n", - "To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.\n", - "You will be able to reuse this secret in all of your notebooks.\n", - "Please note that authentication is recommended but still optional to access public models or datasets.\n", - " warnings.warn(\n" - ] - } - ], - "source": [ - "# Folder in COLAB where snapshots will be saved\n", - "model_files = Path(\"hf_files\").resolve()\n", - "model_files.mkdir(exist_ok=True)\n", - "\n", - "# Download the snapshot and model configuration file\n", - "# This is generic code to download any snapshot from HuggingFace\n", - "# To download DeepLabCut SuperAnimal or Model Zoo models, check\n", - "# out dlclibrary!\n", - "path_model_config = Path(\n", - " huggingface_hub.hf_hub_download(\n", - " \"DeepLabCut/HumanBody\",\n", - " \"rtmpose-x_simcc-body7_pytorch_config.yaml\",\n", - " local_dir=model_files,\n", - " )\n", - ")\n", - "path_snapshot = Path(\n", - " huggingface_hub.hf_hub_download(\n", - " \"DeepLabCut/HumanBody\",\n", - " \"rtmpose-x_simcc-body7.pt\",\n", - " local_dir=model_files,\n", - " )\n", - ")" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Running object detection\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "eEqukXXy0coy" - }, - "source": [ - "We'll now also define some parameters that we'll later use to plot predictions:\n", - "\n", - "- a colormap for the keypoints to plot\n", - "- a colormap for the limbs of the skeleton\n", - "- a skeleton for the model\n" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 1/1 [00:00<00:00, 1.95it/s]\n" + ] }, { - "cell_type": "code", - "execution_count": 4, - "metadata": { - "id": "Tam4rfJK0c_b" - }, - "outputs": [], - "source": [ - "cmap_keypoints = plt.get_cmap(\"rainbow\")\n", - "cmap_skeleton = plt.get_cmap(\"rainbow_r\")\n", - "\n", - "bodyparts2connect = [\n", - " (\"right_ankle\", \"right_knee\"),\n", - " (\"right_knee\", \"right_hip\"),\n", - " (\"left_ankle\", \"left_knee\"),\n", - " (\"left_hip\", \"left_knee\"),\n", - " (\"left_hip\", \"right_hip\"),\n", - " (\"right_shoulder\", \"right_hip\"),\n", - " (\"left_shoulder\", \"left_hip\"),\n", - " (\"left_shoulder\", \"right_shoulder\"),\n", - " (\"left_shoulder\", \"left_elbow\"),\n", - " (\"right_shoulder\", \"right_elbow\"),\n", - " (\"left_elbow\", \"left_wrist\"),\n", - " (\"right_elbow\", \"right_wrist\"),\n", - " (\"right_eye\", \"left_ear\"),\n", - " (\"left_eye\", \"right_eye\"),\n", - " (\"left_eye\", \"left_ear\"),\n", - " (\"right_eye\", \"right_ear\"),\n", - " (\"left_ear\", \"left_shoulder\"),\n", - " (\"right_ear\", \"right_shoulder\"),\n", - " (\"left_shoulder\", \"left_elbow\"),\n", - " (\"right_shoulder\", \"right_elbow\"),\n", - "]\n", - "skeleton = [\n", - " [16, 14],\n", - " [14, 12],\n", - " [17, 15],\n", - " [15, 13],\n", - " [12, 13],\n", - " [6, 12],\n", - " [7, 13],\n", - " [6, 7],\n", - " [6, 8],\n", - " [7, 9],\n", - " [8, 10],\n", - " [9, 11],\n", - " [2, 3],\n", - " [1, 2],\n", - " [1, 3],\n", - " [2, 4],\n", - " [3, 5],\n", - " [4, 6],\n", - " [5, 7],\n", - "]" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Running pose estimation\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "cCxkkd-b0EJq" - }, - "source": [ - "## Running Inference on Images" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "1it [00:00, 78.27it/s]\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "dotn_xN-05gh" - }, - "source": [ - "First, let's upload some images to run inference on. To do so, you can just run the cell below." - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Saving the predictions to a CSV file\n", + "Done!\n" + ] + } + ], + "source": [ + "# Define the device on which the models will run\n", + "device = \"cuda\" # e.g. cuda, cpu\n", + "\n", + "# The maximum number of detections to keep in an image\n", + "max_detections = 10\n", + "\n", + "#############################################\n", + "# Run a pretrained detector to get bounding boxes\n", + "\n", + "# Load the detector from torchvision\n", + "weights = detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights.DEFAULT\n", + "detector = detection.fasterrcnn_mobilenet_v3_large_fpn(\n", + " weights=weights, box_score_thresh=0.6,\n", + ")\n", + "detector.eval()\n", + "detector.to(device)\n", + "preprocess = weights.transforms()\n", + "\n", + "# The context is a list containing the bounding boxes predicted\n", + "# for each image; it will be given to the RTMPose model alongside\n", + "# the images.\n", + "context = []\n", + "\n", + "print(\"Running object detection\")\n", + "with torch.no_grad():\n", + " for image_path in tqdm(image_paths):\n", + " image = Image.open(image_path).convert(\"RGB\")\n", + " batch = [preprocess(image).to(device)]\n", + " predictions = detector(batch)[0]\n", + " bboxes = predictions[\"boxes\"].cpu().numpy()\n", + " labels = predictions[\"labels\"].cpu().numpy()\n", + "\n", + " # Obtain the bounding boxes predicted for humans\n", + " human_bboxes = [\n", + " bbox for bbox, label in zip(bboxes, labels) if label == 1\n", + " ]\n", + "\n", + " # Convert bounding boxes to xywh format\n", + " bboxes = np.zeros((0, 4))\n", + " if len(human_bboxes) > 0:\n", + " bboxes = np.stack(human_bboxes)\n", + " bboxes[:, 2] -= bboxes[:, 0]\n", + " bboxes[:, 3] -= bboxes[:, 1]\n", + "\n", + " # Only keep the best N detections\n", + " bboxes = bboxes[:max_detections]\n", + "\n", + " context.append({\"bboxes\": bboxes})\n", + "\n", + "\n", + "#############################################\n", + "# Run inference on the images\n", + "pose_cfg = dlc_torch.config.read_config_as_dict(path_model_config)\n", + "runner = dlc_torch.get_pose_inference_runner(\n", + " pose_cfg,\n", + " snapshot_path=path_snapshot,\n", + " batch_size=16,\n", + " max_individuals=max_detections,\n", + ")\n", + "\n", + "print(\"Running pose estimation\")\n", + "predictions = runner.inference(tqdm(zip(image_paths, context)))\n", + "\n", + "\n", + "#############################################\n", + "# Create a DataFrame with the predictions, and save them to a CSV file.\n", + "print(\"Saving the predictions to a CSV file\")\n", + "df = dlc_torch.build_predictions_dataframe(\n", + " scorer=\"rtmpose-body7\",\n", + " predictions={\n", + " img_path: img_predictions\n", + " for img_path, img_predictions in zip(image_paths, predictions)\n", + " },\n", + " parameters=dlc_torch.PoseDatasetParameters(\n", + " bodyparts=pose_cfg[\"metadata\"][\"bodyparts\"],\n", + " unique_bpts=pose_cfg[\"metadata\"][\"unique_bodyparts\"],\n", + " individuals=[f\"idv_{i}\" for i in range(max_detections)]\n", + " )\n", + ")\n", + "\n", + "# Save to CSV\n", + "df.to_csv(\"image_predictions.csv\")\n", + "\n", + "print(\"Done!\")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "pWtdL4U52OBJ" + }, + "source": [ + "Finally, we can plot the predictions!" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 447 }, + "id": "3slKu6Lr2MUh", + "outputId": "ef7d938c-39fc-473a-9b88-6169cbfbc567" + }, + "outputs": [ { - "cell_type": "code", - "execution_count": 5, - "metadata": { - "id": "mZtikE1H0D34", - "outputId": "3d47314f-3ed0-40b2-e54d-2677feef9943", - "colab": { - "base_uri": "https://localhost:8080/", - "height": 92 - } - }, - "outputs": [ - { - "output_type": "display_data", - "data": { - "text/plain": [ - "" - ], - "text/html": [ - "\n", - " \n", - " \n", - " Upload widget is only available when the cell has been executed in the\n", - " current browser session. Please rerun this cell to enable.\n", - " \n", - " " - ] - }, - "metadata": {} - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Saving taylor_swift.jpg to taylor_swift.jpg\n", - "User uploaded file 'taylor_swift.jpg' with length 46915 bytes\n" - ] - } - ], - "source": [ - "from google.colab import files\n", - "\n", - "#JPG or PNG is recommended:\n", - "uploaded = files.upload()\n", - "for filepath, content in uploaded.items():\n", - " print(f\"User uploaded file '{filepath}' with length {len(content)} bytes\")\n", - "\n", - "image_paths = [Path(filepath).resolve() for filepath in uploaded.keys()]\n", - "\n", - "# If this cell fails (e.g., when using Safari in place of Google Chrome),\n", - "# manually upload your image via the Files menu to the left and define\n", - "# `image_paths` yourself with right `click` > `copy path` on the image:\n", - "#\n", - "# image_paths = [\n", - "# Path(\"/path/to/my/image_000.png\"),\n", - "# Path(\"/path/to/my/image_001.png\"),\n", - "# ]\n" + "data": { + "image/png": "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\n", + "text/plain": [ + "

" ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "#############################################\n", + "# Unpack and plot predictions\n", + "plot_skeleton = True\n", + "plot_pose_markers = True\n", + "plot_bounding_boxes = True\n", + "marker_size = 12\n", + "\n", + "for image_path, image_predictions in zip(image_paths, predictions):\n", + " image = Image.open(image_path).convert(\"RGB\")\n", + "\n", + " pose = image_predictions[\"bodyparts\"]\n", + " bboxes = image_predictions[\"bboxes\"]\n", + " num_individuals, num_bodyparts = pose.shape[:2]\n", + "\n", + " fig, ax = plt.subplots(figsize=(8, 8))\n", + " ax.imshow(image)\n", + " ax.set_xlim(0, image.width)\n", + " ax.set_ylim(image.height, 0)\n", + " ax.axis(\"off\")\n", + " for idv_pose in pose:\n", + " if plot_skeleton:\n", + " bones = []\n", + " for bpt_1, bpt_2 in skeleton:\n", + " bones.append([idv_pose[bpt_1 - 1, :2], idv_pose[bpt_2 - 1, :2]])\n", + "\n", + " bone_colors = cmap_skeleton\n", + " if not isinstance(cmap_skeleton, str):\n", + " bone_colors = cmap_skeleton(np.linspace(0, 1, len(skeleton)))\n", + "\n", + " ax.add_collection(\n", + " collections.LineCollection(bones, colors=bone_colors)\n", + " )\n", + "\n", + " if plot_pose_markers:\n", + " ax.scatter(\n", + " idv_pose[:, 0],\n", + " idv_pose[:, 1],\n", + " c=list(range(num_bodyparts)),\n", + " cmap=\"rainbow\",\n", + " s=marker_size,\n", + " )\n", + "\n", + " if plot_bounding_boxes:\n", + " for x, y, w, h in bboxes:\n", + " ax.plot(\n", + " [x, x + w, x + w, x, x],\n", + " [y, y, y + h, y + h, y],\n", + " c=\"r\",\n", + " )\n", + "\n", + " plt.show()" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "wO18A_3m5Spk" + }, + "source": [ + "## Running Inference on a Video\n", + "\n", + "Running pose inference on a video is very similar! First, upload a video to Google Drive." + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 92 }, + "id": "d9a7gSe15bCa", + "outputId": "698b180c-cd8f-4d17-9c71-f8e58f93631b" + }, + "outputs": [ { - "cell_type": "code", - "execution_count": 6, - "metadata": { - "id": "nj-HtOBSwtdk", - "outputId": "eb5f3b18-cc89-4dd1-a58e-6c39c62582af", - "colab": { - "base_uri": "https://localhost:8080/" - } - }, - "outputs": [ - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Running object detection\n" - ] - }, - { - "output_type": "stream", - "name": "stderr", - "text": [ - "100%|██████████| 1/1 [00:00<00:00, 1.95it/s]\n" - ] - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Running pose estimation\n" - ] - }, - { - "output_type": "stream", - "name": "stderr", - "text": [ - "1it [00:00, 78.27it/s]\n" - ] - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Saving the predictions to a CSV file\n", - "Done!\n" - ] - } + "data": { + "text/html": [ + "\n", + " \n", + " \n", + " Upload widget is only available when the cell has been executed in the\n", + " current browser session. Please rerun this cell to enable.\n", + " \n", + " " ], - "source": [ - "# Define the device on which the models will run\n", - "device = \"cuda\" # e.g. cuda, cpu\n", - "\n", - "# The maximum number of detections to keep in an image\n", - "max_detections = 10\n", - "\n", - "#############################################\n", - "# Run a pretrained detector to get bounding boxes\n", - "\n", - "# Load the detector from torchvision\n", - "weights = detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights.DEFAULT\n", - "detector = detection.fasterrcnn_mobilenet_v3_large_fpn(\n", - " weights=weights, box_score_thresh=0.6,\n", - ")\n", - "detector.eval()\n", - "detector.to(device)\n", - "preprocess = weights.transforms()\n", - "\n", - "# The context is a list containing the bounding boxes predicted\n", - "# for each image; it will be given to the RTMPose model alongside\n", - "# the images.\n", - "context = []\n", - "\n", - "print(\"Running object detection\")\n", - "with torch.no_grad():\n", - " for image_path in tqdm(image_paths):\n", - " image = Image.open(image_path).convert(\"RGB\")\n", - " batch = [preprocess(image).to(device)]\n", - " predictions = detector(batch)[0]\n", - " bboxes = predictions[\"boxes\"].cpu().numpy()\n", - " labels = predictions[\"labels\"].cpu().numpy()\n", - "\n", - " # Obtain the bounding boxes predicted for humans\n", - " human_bboxes = [\n", - " bbox for bbox, label in zip(bboxes, labels) if label == 1\n", - " ]\n", - "\n", - " # Convert bounding boxes to xywh format\n", - " bboxes = np.zeros((0, 4))\n", - " if len(human_bboxes) > 0:\n", - " bboxes = np.stack(human_bboxes)\n", - " bboxes[:, 2] -= bboxes[:, 0]\n", - " bboxes[:, 3] -= bboxes[:, 1]\n", - "\n", - " # Only keep the best N detections\n", - " bboxes = bboxes[:max_detections]\n", - "\n", - " context.append({\"bboxes\": bboxes})\n", - "\n", - "\n", - "#############################################\n", - "# Run inference on the images\n", - "pose_cfg = dlc_torch.config.read_config_as_dict(path_model_config)\n", - "runner = dlc_torch.get_pose_inference_runner(\n", - " pose_cfg,\n", - " snapshot_path=path_snapshot,\n", - " batch_size=16,\n", - " max_individuals=max_detections,\n", - ")\n", - "\n", - "print(\"Running pose estimation\")\n", - "predictions = runner.inference(tqdm(zip(image_paths, context)))\n", - "\n", - "\n", - "#############################################\n", - "# Create a DataFrame with the predictions, and save them to a CSV file.\n", - "print(\"Saving the predictions to a CSV file\")\n", - "df = dlc_torch.build_predictions_dataframe(\n", - " scorer=\"rtmpose-body7\",\n", - " predictions={\n", - " img_path: img_predictions\n", - " for img_path, img_predictions in zip(image_paths, predictions)\n", - " },\n", - " parameters=dlc_torch.PoseDatasetParameters(\n", - " bodyparts=pose_cfg[\"metadata\"][\"bodyparts\"],\n", - " unique_bpts=pose_cfg[\"metadata\"][\"unique_bodyparts\"],\n", - " individuals=[f\"idv_{i}\" for i in range(max_detections)]\n", - " )\n", - ")\n", - "\n", - "# Save to CSV\n", - "df.to_csv(\"image_predictions.csv\")\n", - "\n", - "print(\"Done!\")" + "text/plain": [ + "" ] + }, + "metadata": {}, + "output_type": "display_data" }, { - "cell_type": "markdown", - "metadata": { - "id": "pWtdL4U52OBJ" - }, - "source": [ - "Finally, we can plot the predictions!" - ] - }, - { - "cell_type": "code", - "execution_count": 7, - "metadata": { - "id": "3slKu6Lr2MUh", - "outputId": "ef7d938c-39fc-473a-9b88-6169cbfbc567", - "colab": { - "base_uri": "https://localhost:8080/", - "height": 447 - } - }, - "outputs": [ - { - "output_type": "display_data", - "data": { - "text/plain": [ - "
" - ], - "image/png": "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\n" - }, - "metadata": {} - } - ], - "source": [ - "#############################################\n", - "# Unpack and plot predictions\n", - "plot_skeleton = True\n", - "plot_pose_markers = True\n", - "plot_bounding_boxes = True\n", - "marker_size = 12\n", - "\n", - "for image_path, image_predictions in zip(image_paths, predictions):\n", - " image = Image.open(image_path).convert(\"RGB\")\n", - "\n", - " pose = image_predictions[\"bodyparts\"]\n", - " bboxes = image_predictions[\"bboxes\"]\n", - " num_individuals, num_bodyparts = pose.shape[:2]\n", - "\n", - " fig, ax = plt.subplots(figsize=(8, 8))\n", - " ax.imshow(image)\n", - " ax.set_xlim(0, image.width)\n", - " ax.set_ylim(image.height, 0)\n", - " ax.axis(\"off\")\n", - " for idv_pose in pose:\n", - " if plot_skeleton:\n", - " bones = []\n", - " for bpt_1, bpt_2 in skeleton:\n", - " bones.append([idv_pose[bpt_1 - 1, :2], idv_pose[bpt_2 - 1, :2]])\n", - "\n", - " bone_colors = cmap_skeleton\n", - " if not isinstance(cmap_skeleton, str):\n", - " bone_colors = cmap_skeleton(np.linspace(0, 1, len(skeleton)))\n", - "\n", - " ax.add_collection(\n", - " collections.LineCollection(bones, colors=bone_colors)\n", - " )\n", - "\n", - " if plot_pose_markers:\n", - " ax.scatter(\n", - " idv_pose[:, 0],\n", - " idv_pose[:, 1],\n", - " c=list(range(num_bodyparts)),\n", - " cmap=\"rainbow\",\n", - " s=marker_size,\n", - " )\n", - "\n", - " if plot_bounding_boxes:\n", - " for x, y, w, h in bboxes:\n", - " ax.plot(\n", - " [x, x + w, x + w, x, x],\n", - " [y, y, y + h, y + h, y],\n", - " c=\"r\",\n", - " )\n", - "\n", - " plt.show()" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Saving taylor-dancing.mov to taylor-dancing.mov\n", + "User uploaded file 'taylor-dancing.mov' with length 1415324 bytes\n" + ] + } + ], + "source": [ + "from google.colab import files\n", + "\n", + "uploaded = files.upload()\n", + "for filepath, content in uploaded.items():\n", + " print(f\"User uploaded file '{filepath}' with length {len(content)} bytes\")\n", + "\n", + "\n", + "video_path = [Path(filepath).resolve() for filepath in uploaded.keys()][0]\n", + "\n", + "# If this cell fails (e.g., when using Safari in place of Google Chrome),\n", + "# manually upload your video via the Files menu to the left and define\n", + "# `video_path` yourself with right `click` > `copy path` on the video:\n", + "#\n", + "# video_path = Path(\"/path/to/my/video.mp4\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "I885B01359qu", + "outputId": "0affdeda-a10b-4849-b3cd-edf1cb202b52" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "id": "wO18A_3m5Spk" - }, - "source": [ - "## Running Inference on a Video\n", - "\n", - "Running pose inference on a video is very similar! First, upload a video to Google Drive." - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Running object detection\n" + ] }, { - "cell_type": "code", - "execution_count": 8, - "metadata": { - "id": "d9a7gSe15bCa", - "outputId": "698b180c-cd8f-4d17-9c71-f8e58f93631b", - "colab": { - "base_uri": "https://localhost:8080/", - "height": 92 - } - }, - "outputs": [ - { - "output_type": "display_data", - "data": { - "text/plain": [ - "" - ], - "text/html": [ - "\n", - " \n", - " \n", - " Upload widget is only available when the cell has been executed in the\n", - " current browser session. Please rerun this cell to enable.\n", - " \n", - " " - ] - }, - "metadata": {} - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Saving taylor-dancing.mov to taylor-dancing.mov\n", - "User uploaded file 'taylor-dancing.mov' with length 1415324 bytes\n" - ] - } - ], - "source": [ - "from google.colab import files\n", - "\n", - "uploaded = files.upload()\n", - "for filepath, content in uploaded.items():\n", - " print(f\"User uploaded file '{filepath}' with length {len(content)} bytes\")\n", - "\n", - "\n", - "video_path = [Path(filepath).resolve() for filepath in uploaded.keys()][0]\n", - "\n", - "# If this cell fails (e.g., when using Safari in place of Google Chrome),\n", - "# manually upload your video via the Files menu to the left and define\n", - "# `video_path` yourself with right `click` > `copy path` on the video:\n", - "#\n", - "# video_path = Path(\"/path/to/my/video.mp4\")\n" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + " 81%|████████▏ | 66/81 [00:02<00:00, 25.37it/s]\n" + ] }, { - "cell_type": "code", - "execution_count": 9, - "metadata": { - "id": "I885B01359qu", - "outputId": "0affdeda-a10b-4849-b3cd-edf1cb202b52", - "colab": { - "base_uri": "https://localhost:8080/" - } - }, - "outputs": [ - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Running object detection\n" - ] - }, - { - "output_type": "stream", - "name": "stderr", - "text": [ - " 81%|████████▏ | 66/81 [00:02<00:00, 25.37it/s]\n" - ] - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Running pose estimation\n" - ] - }, - { - "output_type": "stream", - "name": "stderr", - "text": [ - " 81%|████████▏ | 66/81 [00:01<00:00, 53.25it/s]\n" - ] - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Saving the predictions to a CSV file\n", - "Done!\n" - ] - } - ], - "source": [ - "# Define the device on which the models will run\n", - "device = \"cuda\" # e.g. cuda, cpu\n", - "\n", - "# The maximum number of individuals to detect in an image\n", - "max_detections = 30\n", - "\n", - "\n", - "#############################################\n", - "# Create a video iterator\n", - "video = dlc_torch.VideoIterator(video_path)\n", - "\n", - "\n", - "#############################################\n", - "# Run a pretrained detector to get bounding boxes\n", - "\n", - "# Load the detector from torchvision\n", - "weights = detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights.DEFAULT\n", - "detector = detection.fasterrcnn_mobilenet_v3_large_fpn(\n", - " weights=weights, box_score_thresh=0.6,\n", - ")\n", - "detector.eval()\n", - "detector.to(device)\n", - "preprocess = weights.transforms()\n", - "\n", - "# The context is a list containing the bounding boxes predicted for each frame\n", - "# in the video.\n", - "context = []\n", - "\n", - "print(\"Running object detection\")\n", - "with torch.no_grad():\n", - " for frame in tqdm(video):\n", - " batch = [preprocess(Image.fromarray(frame)).to(device)]\n", - " predictions = detector(batch)[0]\n", - " bboxes = predictions[\"boxes\"].cpu().numpy()\n", - " labels = predictions[\"labels\"].cpu().numpy()\n", - "\n", - " # Obtain the bounding boxes predicted for humans\n", - " human_bboxes = [\n", - " bbox for bbox, label in zip(bboxes, labels) if label == 1\n", - " ]\n", - "\n", - " # Convert bounding boxes to xywh format\n", - " bboxes = np.zeros((0, 4))\n", - " if len(human_bboxes) > 0:\n", - " bboxes = np.stack(human_bboxes)\n", - " bboxes[:, 2] -= bboxes[:, 0]\n", - " bboxes[:, 3] -= bboxes[:, 1]\n", - "\n", - " # Only keep the top N bounding boxes\n", - " bboxes = bboxes[:max_detections]\n", - "\n", - " context.append({\"bboxes\": bboxes})\n", - "\n", - "# Set the context for the video\n", - "video.set_context(context)\n", - "\n", - "\n", - "#############################################\n", - "# Run inference on the images (in this case a single image)\n", - "pose_cfg = dlc_torch.config.read_config_as_dict(path_model_config)\n", - "runner = dlc_torch.get_pose_inference_runner(\n", - " pose_cfg,\n", - " snapshot_path=path_snapshot,\n", - " batch_size=16,\n", - " max_individuals=max_detections,\n", - ")\n", - "\n", - "print(\"Running pose estimation\")\n", - "predictions = runner.inference(tqdm(video))\n", - "\n", - "\n", - "print(\"Saving the predictions to a CSV file\")\n", - "df = dlc_torch.build_predictions_dataframe(\n", - " scorer=\"rtmpose-body7\",\n", - " predictions={\n", - " idx: img_predictions\n", - " for idx, img_predictions in enumerate(predictions)\n", - " },\n", - " parameters=dlc_torch.PoseDatasetParameters(\n", - " bodyparts=pose_cfg[\"metadata\"][\"bodyparts\"],\n", - " unique_bpts=pose_cfg[\"metadata\"][\"unique_bodyparts\"],\n", - " individuals=[f\"idv_{i}\" for i in range(max_detections)]\n", - " )\n", - ")\n", - "df.to_csv(\"video_predictions.csv\")\n", - "\n", - "print(\"Done!\")" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Running pose estimation\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "altka3NGB_su" - }, - "source": [ - "Finally, we can plot the predictions on the video! The labeled video output is saved in the `\"video_predictions.mp4\"` file, and can be downloaded to be viewed." - ] + "name": "stderr", + "output_type": "stream", + "text": [ + " 81%|████████▏ | 66/81 [00:01<00:00, 53.25it/s]\n" + ] }, { - "cell_type": "code", - "execution_count": 10, - "metadata": { - "id": "xRWxH0gO6oPg", - "outputId": "c2cc9025-7741-4403-d5cc-c62470a4ba74", - "colab": { - "base_uri": "https://localhost:8080/" - } - }, - "outputs": [ - { - "output_type": "stream", - "name": "stderr", - "text": [ - "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/make_labeled_video.py:146: FutureWarning: DataFrame.groupby with axis=1 is deprecated. Do `frame.T.groupby(...)` without axis instead.\n", - " Dataframe.groupby(level=\"individuals\", axis=1).size().values // 3\n" - ] - }, - { - "output_type": "stream", - "name": "stdout", - "text": [ - "Duration of video [s]: 1.57, recorded with 51.7 fps!\n", - "Overall # of frames: 81 with cropped frame dimensions: 828 768\n", - "Generating frames and creating video.\n" - ] - }, - { - "output_type": "stream", - "name": "stderr", - "text": [ - "100%|██████████| 66/66 [00:01<00:00, 35.27it/s]\n" - ] - } - ], - "source": [ - "from deeplabcut.utils.make_labeled_video import CreateVideo\n", - "from deeplabcut.utils.video_processor import VideoProcessorCV\n", - "\n", - "video_output_path = \"video_predictions.mp4\"\n", - "\n", - "clip = VideoProcessorCV(str(video_path), sname=video_output_path, codec=\"mp4v\")\n", - "CreateVideo(\n", - " clip,\n", - " df,\n", - " pcutoff=0.4,\n", - " dotsize=3,\n", - " colormap=\"rainbow\",\n", - " bodyparts2plot=pose_cfg[\"metadata\"][\"bodyparts\"],\n", - " trailpoints=0,\n", - " cropping=False,\n", - " x1=0,\n", - " x2=clip.w,\n", - " y1=0,\n", - " y2=clip.h,\n", - " bodyparts2connect=bodyparts2connect,\n", - " skeleton_color=\"w\",\n", - " draw_skeleton=True,\n", - " displaycropped=True,\n", - " color_by=\"bodypart\",\n", - ")" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Saving the predictions to a CSV file\n", + "Done!\n" + ] } - ], - "metadata": { - "accelerator": "GPU", + ], + "source": [ + "# Define the device on which the models will run\n", + "device = \"cuda\" # e.g. cuda, cpu\n", + "\n", + "# The maximum number of individuals to detect in an image\n", + "max_detections = 30\n", + "\n", + "\n", + "#############################################\n", + "# Create a video iterator\n", + "video = dlc_torch.VideoIterator(video_path)\n", + "\n", + "\n", + "#############################################\n", + "# Run a pretrained detector to get bounding boxes\n", + "\n", + "# Load the detector from torchvision\n", + "weights = detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights.DEFAULT\n", + "detector = detection.fasterrcnn_mobilenet_v3_large_fpn(\n", + " weights=weights, box_score_thresh=0.6,\n", + ")\n", + "detector.eval()\n", + "detector.to(device)\n", + "preprocess = weights.transforms()\n", + "\n", + "# The context is a list containing the bounding boxes predicted for each frame\n", + "# in the video.\n", + "context = []\n", + "\n", + "print(\"Running object detection\")\n", + "with torch.no_grad():\n", + " for frame in tqdm(video):\n", + " batch = [preprocess(Image.fromarray(frame)).to(device)]\n", + " predictions = detector(batch)[0]\n", + " bboxes = predictions[\"boxes\"].cpu().numpy()\n", + " labels = predictions[\"labels\"].cpu().numpy()\n", + "\n", + " # Obtain the bounding boxes predicted for humans\n", + " human_bboxes = [\n", + " bbox for bbox, label in zip(bboxes, labels) if label == 1\n", + " ]\n", + "\n", + " # Convert bounding boxes to xywh format\n", + " bboxes = np.zeros((0, 4))\n", + " if len(human_bboxes) > 0:\n", + " bboxes = np.stack(human_bboxes)\n", + " bboxes[:, 2] -= bboxes[:, 0]\n", + " bboxes[:, 3] -= bboxes[:, 1]\n", + "\n", + " # Only keep the top N bounding boxes\n", + " bboxes = bboxes[:max_detections]\n", + "\n", + " context.append({\"bboxes\": bboxes})\n", + "\n", + "# Set the context for the video\n", + "video.set_context(context)\n", + "\n", + "\n", + "#############################################\n", + "# Run inference on the images (in this case a single image)\n", + "pose_cfg = dlc_torch.config.read_config_as_dict(path_model_config)\n", + "runner = dlc_torch.get_pose_inference_runner(\n", + " pose_cfg,\n", + " snapshot_path=path_snapshot,\n", + " batch_size=16,\n", + " max_individuals=max_detections,\n", + ")\n", + "\n", + "print(\"Running pose estimation\")\n", + "predictions = runner.inference(tqdm(video))\n", + "\n", + "\n", + "print(\"Saving the predictions to a CSV file\")\n", + "df = dlc_torch.build_predictions_dataframe(\n", + " scorer=\"rtmpose-body7\",\n", + " predictions={\n", + " idx: img_predictions\n", + " for idx, img_predictions in enumerate(predictions)\n", + " },\n", + " parameters=dlc_torch.PoseDatasetParameters(\n", + " bodyparts=pose_cfg[\"metadata\"][\"bodyparts\"],\n", + " unique_bpts=pose_cfg[\"metadata\"][\"unique_bodyparts\"],\n", + " individuals=[f\"idv_{i}\" for i in range(max_detections)]\n", + " )\n", + ")\n", + "df.to_csv(\"video_predictions.csv\")\n", + "\n", + "print(\"Done!\")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "altka3NGB_su" + }, + "source": [ + "Finally, we can plot the predictions on the video! The labeled video output is saved in the `\"video_predictions.mp4\"` file, and can be downloaded to be viewed." + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": { "colab": { - "gpuType": "T4", - "provenance": [], - "include_colab_link": true + "base_uri": "https://localhost:8080/" }, - "kernelspec": { - "display_name": "Python 3", - "name": "python3" + "id": "xRWxH0gO6oPg", + "outputId": "c2cc9025-7741-4403-d5cc-c62470a4ba74" + }, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/make_labeled_video.py:146: FutureWarning: DataFrame.groupby with axis=1 is deprecated. Do `frame.T.groupby(...)` without axis instead.\n", + " Dataframe.groupby(level=\"individuals\", axis=1).size().values // 3\n" + ] }, - "language_info": { - "name": "python" + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Duration of video [s]: 1.57, recorded with 51.7 fps!\n", + "Overall # of frames: 81 with cropped frame dimensions: 828 768\n", + "Generating frames and creating video.\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 66/66 [00:01<00:00, 35.27it/s]\n" + ] } + ], + "source": [ + "from deeplabcut.utils.make_labeled_video import CreateVideo\n", + "from deeplabcut.utils.video_processor import VideoProcessorCV\n", + "\n", + "video_output_path = \"video_predictions.mp4\"\n", + "\n", + "clip = VideoProcessorCV(str(video_path), sname=video_output_path, codec=\"mp4v\")\n", + "CreateVideo(\n", + " clip,\n", + " df,\n", + " pcutoff=0.4,\n", + " dotsize=3,\n", + " colormap=\"rainbow\",\n", + " bodyparts2plot=pose_cfg[\"metadata\"][\"bodyparts\"],\n", + " trailpoints=0,\n", + " cropping=False,\n", + " x1=0,\n", + " x2=clip.w,\n", + " y1=0,\n", + " y2=clip.h,\n", + " bodyparts2connect=bodyparts2connect,\n", + " skeleton_color=\"w\",\n", + " draw_skeleton=True,\n", + " displaycropped=True,\n", + " color_by=\"bodypart\",\n", + ")" + ] + } + ], + "metadata": { + "accelerator": "GPU", + "colab": { + "gpuType": "T4", + "include_colab_link": true, + "provenance": [] + }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-07-07", + "last_metadata_updated": "2026-03-06" + }, + "kernelspec": { + "display_name": "Python 3", + "name": "python3" }, - "nbformat": 4, - "nbformat_minor": 0 + "language_info": { + "name": "python" + } + }, + "nbformat": 4, + "nbformat_minor": 0 } diff --git a/examples/COLAB/COLAB_YOURDATA_SuperAnimal.ipynb b/examples/COLAB/COLAB_YOURDATA_SuperAnimal.ipynb index c4d1fc575e..f1f0150792 100644 --- a/examples/COLAB/COLAB_YOURDATA_SuperAnimal.ipynb +++ b/examples/COLAB/COLAB_YOURDATA_SuperAnimal.ipynb @@ -1165,6 +1165,11 @@ "gpuType": "T4", "provenance": [] }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-10", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/examples/COLAB/COLAB_YOURDATA_TrainNetwork_VideoAnalysis.ipynb b/examples/COLAB/COLAB_YOURDATA_TrainNetwork_VideoAnalysis.ipynb index 71e4e88045..ed04c2f560 100644 --- a/examples/COLAB/COLAB_YOURDATA_TrainNetwork_VideoAnalysis.ipynb +++ b/examples/COLAB/COLAB_YOURDATA_TrainNetwork_VideoAnalysis.ipynb @@ -414,6 +414,11 @@ "provenance": [], "toc_visible": true }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-16", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3.8.12 ('dlc')", "language": "python", diff --git a/examples/COLAB/COLAB_YOURDATA_maDLC_TrainNetwork_VideoAnalysis.ipynb b/examples/COLAB/COLAB_YOURDATA_maDLC_TrainNetwork_VideoAnalysis.ipynb index b8c060960d..08633839fb 100644 --- a/examples/COLAB/COLAB_YOURDATA_maDLC_TrainNetwork_VideoAnalysis.ipynb +++ b/examples/COLAB/COLAB_YOURDATA_maDLC_TrainNetwork_VideoAnalysis.ipynb @@ -529,6 +529,11 @@ "name": "COLAB_maDLC_TrainNetwork_VideoAnalysis.ipynb", "provenance": [] }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2026-02-10", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3", "name": "python3" diff --git a/examples/COLAB/COLAB_transformer_reID.ipynb b/examples/COLAB/COLAB_transformer_reID.ipynb index 008255692f..e0b222627a 100644 --- a/examples/COLAB/COLAB_transformer_reID.ipynb +++ b/examples/COLAB/COLAB_transformer_reID.ipynb @@ -1,634 +1,639 @@ { - "cells": [ + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "colab_type": "text", + "id": "view-in-github" + }, + "source": [ + "\"Open" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "TGChzLdc-lUJ" + }, + "source": [ + "# Demo: How to use our Pose Transformer for unsupervised identity tracking of animals\n", + "![alt text](https://images.squarespace-cdn.com/content/v1/57f6d51c9f74566f55ecf271/1628250004229-KVYD7JJVHYEFDJ32L9VJ/DLClogo2021.jpg?format=1000w)\n", + "\n", + "https://github.com/DeepLabCut/DeepLabCut\n", + "\n", + "### This notebook illustrates how to use the transformer for a multi-animal DeepLabCut (maDLC) Demo tri-mouse project:\n", + "- load our mini-demo data that includes a pretrained model and unlabeled video.\n", + "- analyze a novel video.\n", + "- use the transformer to do unsupervised ID tracking.\n", + "- create quality check plots and video.\n", + "\n", + "### To create a full maDLC pipeline please see our full docs: https://deeplabcut.github.io/DeepLabCut/README.html\n", + "- Of interest is a full how-to for maDLC: https://deeplabcut.github.io/DeepLabCut/docs/maDLC_UserGuide.html\n", + "- a quick guide to maDLC: https://deeplabcut.github.io/DeepLabCut/docs/quick-start/tutorial_maDLC.html\n", + "- a demo COLAB for how to use maDLC on your own data: https://github.com/DeepLabCut/DeepLabCut/blob/main/examples/COLAB/COLAB_YOURDATA_maDLC_TrainNetwork_VideoAnalysis.ipynb\n", + "\n", + "### To get started, please go to \"Runtime\" ->\"change runtime type\"->select \"Python3\", and then select \"GPU\"\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "xOe2hvy85EVP" + }, + "source": [ + "‼️ **Attention: this demo is for maDLC, which is version 2.2**\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "NXmLeZBX45Oe" + }, + "outputs": [], + "source": [ + "# Install DLC version 2.2-2.3 (pre DLC3):\n", + "!pip install \"deeplabcut[tf]\"" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": { + "id": "TlhrVFKN8euh" + }, + "outputs": [], + "source": [ + "import deeplabcut\n", + "import os" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Wid0GTGMAEnZ" + }, + "source": [ + "## Important - Restart the Runtime for the updated packages to be imported!\n", + "\n", + "PLEASE, click \"restart runtime\" from the output above before proceeding!\n", + "\n", + "No information needs edited in the cells below, you can simply click run on each:\n", + "\n", + "### Download our Demo Project from our server:" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "PusLdqbqJi60", + "outputId": "dbe30821-d3a7-443f-de74-6cb0bee49aac" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "colab_type": "text", - "id": "view-in-github" - }, - "source": [ - "\"Open" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Downloading demo-me-2021-07-14.zip...\n" + ] + } + ], + "source": [ + "# Download our demo project:\n", + "import requests\n", + "from io import BytesIO\n", + "from zipfile import ZipFile\n", + "\n", + "url_record = \"https://zenodo.org/api/records/7883589\"\n", + "response = requests.get(url_record)\n", + "if response.status_code == 200:\n", + " file = response.json()[\"files\"][0]\n", + " title = file[\"key\"]\n", + " print(f\"Downloading {title}...\")\n", + " with requests.get(file[\"links\"][\"self\"], stream=True) as r:\n", + " with ZipFile(BytesIO(r.content)) as zf:\n", + " zf.extractall(path=\"/content\")\n", + "else:\n", + " raise ValueError(f\"The URL {url_record} could not be reached.\")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "8iXtySnQB0BE" + }, + "source": [ + "## Analyze a novel 3 mouse video with our maDLC DLCRNet, pretrained on 3 mice data\n", + "\n", + "In one step, since `auto_track=True` you extract detections and association costs, create tracklets, & stitch them. We can use this to compare to the transformer-guided tracking below.\n" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": { + "id": "odYrU3o8BSAr" + }, + "outputs": [], + "source": [ + "project_path = \"/content/demo-me-2021-07-14\"\n", + "config_path = os.path.join(project_path, \"config.yaml\")\n", + "video = os.path.join(project_path, \"videos\", \"videocompressed1.mp4\")" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 520 }, + "id": "U_351Hkv81X-", + "outputId": "f7c30461-101f-47b6-c04f-15809aa5a4bb" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "id": "TGChzLdc-lUJ" - }, - "source": [ - "# Demo: How to use our Pose Transformer for unsupervised identity tracking of animals\n", - "![alt text](https://images.squarespace-cdn.com/content/v1/57f6d51c9f74566f55ecf271/1628250004229-KVYD7JJVHYEFDJ32L9VJ/DLClogo2021.jpg?format=1000w)\n", - "\n", - "https://github.com/DeepLabCut/DeepLabCut\n", - "\n", - "### This notebook illustrates how to use the transformer for a multi-animal DeepLabCut (maDLC) Demo tri-mouse project:\n", - "- load our mini-demo data that includes a pretrained model and unlabeled video.\n", - "- analyze a novel video.\n", - "- use the transformer to do unsupervised ID tracking.\n", - "- create quality check plots and video.\n", - "\n", - "### To create a full maDLC pipeline please see our full docs: https://deeplabcut.github.io/DeepLabCut/README.html\n", - "- Of interest is a full how-to for maDLC: https://deeplabcut.github.io/DeepLabCut/docs/maDLC_UserGuide.html\n", - "- a quick guide to maDLC: https://deeplabcut.github.io/DeepLabCut/docs/quick-start/tutorial_maDLC.html\n", - "- a demo COLAB for how to use maDLC on your own data: https://github.com/DeepLabCut/DeepLabCut/blob/main/examples/COLAB/COLAB_YOURDATA_maDLC_TrainNetwork_VideoAnalysis.ipynb\n", - "\n", - "### To get started, please go to \"Runtime\" ->\"change runtime type\"->select \"Python3\", and then select \"GPU\"\n" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Using snapshot-20000 for model /content/demo-me-2021-07-14/dlc-models/iteration-0/demoJul14-trainset95shuffle0\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "xOe2hvy85EVP" - }, - "source": [ - "‼️ **Attention: this demo is for maDLC, which is version 2.2**\n" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", + " warnings.warn('`layer.apply` is deprecated and '\n" + ] }, { - "cell_type": "code", - "execution_count": null, - "metadata": { - "id": "NXmLeZBX45Oe" - }, - "outputs": [], - "source": [ - "# Install DLC version 2.2-2.3 (pre DLC3):\n", - "!pip install \"deeplabcut[tf]\"" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Activating extracting of PAFs\n", + "Starting to analyze % /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Duration of video [s]: 77.67 , recorded with 30.0 fps!\n", + "Overall # of frames: 2330 found with (before cropping) frame dimensions: 640 480\n", + "Starting to extract posture from the video(s) with batchsize: 8\n" + ] }, { - "cell_type": "code", - "execution_count": 3, - "metadata": { - "id": "TlhrVFKN8euh" - }, - "outputs": [], - "source": [ - "import deeplabcut\n", - "import os" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 2330/2330 [00:39<00:00, 58.83it/s]\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "Wid0GTGMAEnZ" - }, - "source": [ - "## Important - Restart the Runtime for the updated packages to be imported!\n", - "\n", - "PLEASE, click \"restart runtime\" from the output above before proceeding!\n", - "\n", - "No information needs edited in the cells below, you can simply click run on each:\n", - "\n", - "### Download our Demo Project from our server:" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Video Analyzed. Saving results in /content/demo-me-2021-07-14/videos...\n" + ] }, { - "cell_type": "code", - "execution_count": 5, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/" - }, - "id": "PusLdqbqJi60", - "outputId": "dbe30821-d3a7-443f-de74-6cb0bee49aac" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Downloading demo-me-2021-07-14.zip...\n" - ] - } - ], - "source": [ - "# Download our demo project:\n", - "import requests\n", - "from io import BytesIO\n", - "from zipfile import ZipFile\n", - "\n", - "url_record = \"https://zenodo.org/api/records/7883589\"\n", - "response = requests.get(url_record)\n", - "if response.status_code == 200:\n", - " file = response.json()[\"files\"][0]\n", - " title = file[\"key\"]\n", - " print(f\"Downloading {title}...\")\n", - " with requests.get(file[\"links\"][\"self\"], stream=True) as r:\n", - " with ZipFile(BytesIO(r.content)) as zf:\n", - " zf.extractall(path=\"/content\")\n", - "else:\n", - " raise ValueError(f\"The URL {url_record} could not be reached.\")" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/auxfun_multianimal.py:83: UserWarning: default_track_method` is undefined in the config.yaml file and will be set to `ellipse`.\n", + " warnings.warn(\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "8iXtySnQB0BE" - }, - "source": [ - "## Analyze a novel 3 mouse video with our maDLC DLCRNet, pretrained on 3 mice data\n", - "\n", - "In one step, since `auto_track=True` you extract detections and association costs, create tracklets, & stitch them. We can use this to compare to the transformer-guided tracking below.\n" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Using snapshot-20000 for model /content/demo-me-2021-07-14/dlc-models/iteration-0/demoJul14-trainset95shuffle0\n", + "Processing... /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Analyzing /content/demo-me-2021-07-14/videos/videocompressed1DLC_dlcrnetms5_demoJul14shuffle0_20000.h5\n" + ] }, { - "cell_type": "code", - "execution_count": 6, - "metadata": { - "id": "odYrU3o8BSAr" - }, - "outputs": [], - "source": [ - "project_path = \"/content/demo-me-2021-07-14\"\n", - "config_path = os.path.join(project_path, \"config.yaml\")\n", - "video = os.path.join(project_path, \"videos\", \"videocompressed1.mp4\")" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 2330/2330 [00:02<00:00, 1088.72it/s]\n", + "2330it [00:06, 342.29it/s] \n" + ] }, { - "cell_type": "code", - "execution_count": 7, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/", - "height": 520 - }, - "id": "U_351Hkv81X-", - "outputId": "f7c30461-101f-47b6-c04f-15809aa5a4bb" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Using snapshot-20000 for model /content/demo-me-2021-07-14/dlc-models/iteration-0/demoJul14-trainset95shuffle0\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", - " warnings.warn('`layer.apply` is deprecated and '\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Activating extracting of PAFs\n", - "Starting to analyze % /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Duration of video [s]: 77.67 , recorded with 30.0 fps!\n", - "Overall # of frames: 2330 found with (before cropping) frame dimensions: 640 480\n", - "Starting to extract posture from the video(s) with batchsize: 8\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "100%|██████████| 2330/2330 [00:39<00:00, 58.83it/s]\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Video Analyzed. Saving results in /content/demo-me-2021-07-14/videos...\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/auxfun_multianimal.py:83: UserWarning: default_track_method` is undefined in the config.yaml file and will be set to `ellipse`.\n", - " warnings.warn(\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Using snapshot-20000 for model /content/demo-me-2021-07-14/dlc-models/iteration-0/demoJul14-trainset95shuffle0\n", - "Processing... /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Analyzing /content/demo-me-2021-07-14/videos/videocompressed1DLC_dlcrnetms5_demoJul14shuffle0_20000.h5\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "100%|██████████| 2330/2330 [00:02<00:00, 1088.72it/s]\n", - "2330it [00:06, 342.29it/s] \n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "The tracklets were created (i.e., under the hood deeplabcut.convert_detections2tracklets was run). Now you can 'refine_tracklets' in the GUI, or run 'deeplabcut.stitch_tracklets'.\n", - "Processing... /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "100%|██████████| 4/4 [00:00<00:00, 1488.53it/s]\n", - "/usr/local/lib/python3.11/dist-packages/deeplabcut/refine_training_dataset/stitch.py:934: FutureWarning: Starting with pandas version 3.0 all arguments of to_hdf except for the argument 'path_or_buf' will be keyword-only.\n", - " df.to_hdf(output_name, \"tracks\", format=\"table\", mode=\"w\")\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "The videos are analyzed. Time to assemble animals and track 'em... \n", - " Call 'create_video_with_all_detections' to check multi-animal detection quality before tracking.\n", - "If the tracking is not satisfactory for some videos, consider expanding the training set. You can use the function 'extract_outlier_frames' to extract a few representative outlier frames.\n" - ] - }, - { - "data": { - "application/vnd.google.colaboratory.intrinsic+json": { - "type": "string" - }, - "text/plain": [ - "'DLC_dlcrnetms5_demoJul14shuffle0_20000'" - ] - }, - "execution_count": 7, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "deeplabcut.analyze_videos(config_path,[video],\n", - " shuffle=0, videotype=\"mp4\",\n", - " auto_track=True)" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "The tracklets were created (i.e., under the hood deeplabcut.convert_detections2tracklets was run). Now you can 'refine_tracklets' in the GUI, or run 'deeplabcut.stitch_tracklets'.\n", + "Processing... /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "zmdSLRTOER00" - }, - "source": [ - "### Next, you compute the local, spatio-temporal grouping and track body part assemblies frame-by-frame:" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 4/4 [00:00<00:00, 1488.53it/s]\n", + "/usr/local/lib/python3.11/dist-packages/deeplabcut/refine_training_dataset/stitch.py:934: FutureWarning: Starting with pandas version 3.0 all arguments of to_hdf except for the argument 'path_or_buf' will be keyword-only.\n", + " df.to_hdf(output_name, \"tracks\", format=\"table\", mode=\"w\")\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "F-d6kXqnGeUP" - }, - "source": [ - "## Create a pretty video output:" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "The videos are analyzed. Time to assemble animals and track 'em... \n", + " Call 'create_video_with_all_detections' to check multi-animal detection quality before tracking.\n", + "If the tracking is not satisfactory for some videos, consider expanding the training set. You can use the function 'extract_outlier_frames' to extract a few representative outlier frames.\n" + ] }, { - "cell_type": "code", - "execution_count": 8, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/" - }, - "id": "aTRbuUQ1FBO0", - "outputId": "0d182f64-512d-463d-a997-226c7199b724" + "data": { + "application/vnd.google.colaboratory.intrinsic+json": { + "type": "string" }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Filtering with median model /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Saving filtered csv poses!\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/usr/local/lib/python3.11/dist-packages/deeplabcut/post_processing/filtering.py:298: FutureWarning: Starting with pandas version 3.0 all arguments of to_hdf except for the argument 'path_or_buf' will be keyword-only.\n", - " data.to_hdf(outdataname, \"df_with_missing\", format=\"table\", mode=\"w\")\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Starting to process video: /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", - "Duration of video [s]: 77.67, recorded with 30.0 fps!\n", - "Overall # of frames: 2330 with cropped frame dimensions: 640 480\n", - "Generating frames and creating video.\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/make_labeled_video.py:140: FutureWarning: DataFrame.groupby with axis=1 is deprecated. Do `frame.T.groupby(...)` without axis instead.\n", - " Dataframe.groupby(level=\"individuals\", axis=1).size().values // 3\n", - "100%|██████████| 2330/2330 [00:31<00:00, 73.04it/s]\n" - ] - }, - { - "data": { - "text/plain": [ - "[True]" - ] - }, - "execution_count": 8, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "#Filter the predictions to remove small jitter, if desired:\n", - "deeplabcut.filterpredictions(config_path, [video], shuffle=0, videotype=\"mp4\")\n", - "deeplabcut.create_labeled_video(\n", - " config_path,\n", - " [video],\n", - " videotype=\"mp4\",\n", - " shuffle=0,\n", - " color_by=\"individual\",\n", - " keypoints_only=False,\n", - " draw_skeleton=True,\n", - " filtered=True,\n", - ")" + "text/plain": [ + "'DLC_dlcrnetms5_demoJul14shuffle0_20000'" ] + }, + "execution_count": 7, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "deeplabcut.analyze_videos(config_path,[video],\n", + " shuffle=0, videotype=\"mp4\",\n", + " auto_track=True)" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "zmdSLRTOER00" + }, + "source": [ + "### Next, you compute the local, spatio-temporal grouping and track body part assemblies frame-by-frame:" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "F-d6kXqnGeUP" + }, + "source": [ + "## Create a pretty video output:" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "aTRbuUQ1FBO0", + "outputId": "0d182f64-512d-463d-a997-226c7199b724" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "id": "AYNlrgeNUG4U" - }, - "source": [ - "Now, on the left panel if you click the folder icon, you will see the project folder \"demo-me..\"; click on this and go into \"videos\" and you can find the \"..._id_labeled.mp4\" video, which you can double-click on to download and inspect!" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Filtering with median model /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Saving filtered csv poses!\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "n7GWMBJUA9x5" - }, - "source": [ - "### Create Plots of your data:\n", - "\n", - "> after running, you can look in \"videos\", \"plot-poses\" to check out the trajectories! (sometimes you need to click the folder refresh icon to see it). Within the folder, for example, see plotmus1.png to vide the bodyparts over time vs. pixel position.\n", - "\n" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/deeplabcut/post_processing/filtering.py:298: FutureWarning: Starting with pandas version 3.0 all arguments of to_hdf except for the argument 'path_or_buf' will be keyword-only.\n", + " data.to_hdf(outdataname, \"df_with_missing\", format=\"table\", mode=\"w\")\n" + ] }, { - "cell_type": "code", - "execution_count": 9, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/" - }, - "id": "7w9BDIA7BB_i", - "outputId": "a163087d-cbcb-4e4d-f461-2e24ed19a80b" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", - "Plots created! Please check the directory \"plot-poses\" within the video directory\n" - ] - } - ], - "source": [ - "deeplabcut.plot_trajectories(config_path, [video], shuffle=0,videotype=\"mp4\")" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Starting to process video: /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", + "Duration of video [s]: 77.67, recorded with 30.0 fps!\n", + "Overall # of frames: 2330 with cropped frame dimensions: 640 480\n", + "Generating frames and creating video.\n" + ] }, { - "cell_type": "markdown", - "metadata": { - "id": "l7BJQq7nxHVz" - }, - "source": [ - "# Transformer for reID\n", - "\n", - "while the tracking here is very good without using the transformer, we want to demo the workflow for you!" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/make_labeled_video.py:140: FutureWarning: DataFrame.groupby with axis=1 is deprecated. Do `frame.T.groupby(...)` without axis instead.\n", + " Dataframe.groupby(level=\"individuals\", axis=1).size().values // 3\n", + "100%|██████████| 2330/2330 [00:31<00:00, 73.04it/s]\n" + ] }, { - "cell_type": "code", - "execution_count": 10, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/" - }, - "id": "5xlO6TVYxQWc", - "outputId": "a433221f-0390-4028-fe68-be0b90adad48" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Using snapshot-20000 for model /content/demo-me-2021-07-14/dlc-models/iteration-0/demoJul14-trainset95shuffle0\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", - " warnings.warn('`layer.apply` is deprecated and '\n", - "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", - " warnings.warn('`layer.apply` is deprecated and '\n", - "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", - " warnings.warn('`layer.apply` is deprecated and '\n", - "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", - " warnings.warn('`layer.apply` is deprecated and '\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Activating extracting of PAFs\n", - "Starting to analyze % /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Duration of video [s]: 77.67 , recorded with 30.0 fps!\n", - "Overall # of frames: 2330 found with (before cropping) frame dimensions: 640 480\n", - "Starting to extract posture\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "100%|██████████| 2330/2330 [01:18<00:00, 29.78it/s]\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "If the tracking is not satisfactory for some videos, consider expanding the training set. You can use the function 'extract_outlier_frames' to extract a few representative outlier frames.\n", - "Epoch 10, train acc: 0.61\n", - "Epoch 10, test acc 0.45\n", - "Epoch 20, train acc: 0.74\n", - "Epoch 20, test acc 0.65\n", - "Epoch 30, train acc: 0.78\n", - "Epoch 30, test acc 0.55\n", - "Epoch 40, train acc: 0.76\n", - "Epoch 40, test acc 0.50\n", - "Epoch 50, train acc: 0.85\n", - "Epoch 50, test acc 0.55\n", - "Epoch 60, train acc: 0.84\n", - "Epoch 60, test acc 0.60\n", - "Epoch 70, train acc: 0.85\n", - "Epoch 70, test acc 0.55\n", - "Epoch 80, train acc: 0.79\n", - "Epoch 80, test acc 0.55\n", - "Epoch 90, train acc: 0.88\n", - "Epoch 90, test acc 0.55\n", - "Epoch 100, train acc: 0.84\n", - "Epoch 100, test acc 0.55\n", - "loading params\n", - "Processing... /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "100%|██████████| 4/4 [00:00<00:00, 483.21it/s]\n", - "/usr/local/lib/python3.11/dist-packages/deeplabcut/refine_training_dataset/stitch.py:934: FutureWarning: Starting with pandas version 3.0 all arguments of to_hdf except for the argument 'path_or_buf' will be keyword-only.\n", - " df.to_hdf(output_name, \"tracks\", format=\"table\", mode=\"w\")\n" - ] - } - ], - "source": [ - "deeplabcut.transformer_reID(\n", - " config_path,\n", - " [video],\n", - " shuffle=0,\n", - " videotype=\"mp4\",\n", - " track_method=\"ellipse\",\n", - " n_triplets=100,\n", - ")" + "data": { + "text/plain": [ + "[True]" ] + }, + "execution_count": 8, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "#Filter the predictions to remove small jitter, if desired:\n", + "deeplabcut.filterpredictions(config_path, [video], shuffle=0, videotype=\"mp4\")\n", + "deeplabcut.create_labeled_video(\n", + " config_path,\n", + " [video],\n", + " videotype=\"mp4\",\n", + " shuffle=0,\n", + " color_by=\"individual\",\n", + " keypoints_only=False,\n", + " draw_skeleton=True,\n", + " filtered=True,\n", + ")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "AYNlrgeNUG4U" + }, + "source": [ + "Now, on the left panel if you click the folder icon, you will see the project folder \"demo-me..\"; click on this and go into \"videos\" and you can find the \"..._id_labeled.mp4\" video, which you can double-click on to download and inspect!" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "n7GWMBJUA9x5" + }, + "source": [ + "### Create Plots of your data:\n", + "\n", + "> after running, you can look in \"videos\", \"plot-poses\" to check out the trajectories! (sometimes you need to click the folder refresh icon to see it). Within the folder, for example, see plotmus1.png to vide the bodyparts over time vs. pixel position.\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "7w9BDIA7BB_i", + "outputId": "a163087d-cbcb-4e4d-f461-2e24ed19a80b" + }, + "outputs": [ { - "cell_type": "markdown", - "metadata": { - "id": "uO_yoqN7xiBT" - }, - "source": [ - "now we can make another video with the transformer-guided tracking:\n" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", + "Plots created! Please check the directory \"plot-poses\" within the video directory\n" + ] + } + ], + "source": [ + "deeplabcut.plot_trajectories(config_path, [video], shuffle=0,videotype=\"mp4\")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "l7BJQq7nxHVz" + }, + "source": [ + "# Transformer for reID\n", + "\n", + "while the tracking here is very good without using the transformer, we want to demo the workflow for you!" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" }, + "id": "5xlO6TVYxQWc", + "outputId": "a433221f-0390-4028-fe68-be0b90adad48" + }, + "outputs": [ { - "cell_type": "code", - "execution_count": 11, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/" - }, - "id": "MBMbRFEMxmi4", - "outputId": "5ca4357a-c8e1-46c6-ecad-141bfce48cc5" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", - "Plots created! Please check the directory \"plot-poses\" within the video directory\n" - ] - } - ], - "source": [ - "deeplabcut.plot_trajectories(\n", - " config_path,\n", - " [video],\n", - " shuffle=0,\n", - " videotype=\"mp4\",\n", - " track_method=\"transformer\",\n", - ")" - ] + "name": "stdout", + "output_type": "stream", + "text": [ + "Using snapshot-20000 for model /content/demo-me-2021-07-14/dlc-models/iteration-0/demoJul14-trainset95shuffle0\n" + ] }, { - "cell_type": "code", - "execution_count": 12, - "metadata": { - "colab": { - "base_uri": "https://localhost:8080/" - }, - "id": "vx3e-r1CoXaX", - "outputId": "46cdbd39-d1f6-4b78-abba-7e979740f2a2" - }, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Starting to process video: /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", - "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", - "Duration of video [s]: 77.67, recorded with 30.0 fps!\n", - "Overall # of frames: 2330 with cropped frame dimensions: 640 480\n", - "Generating frames and creating video.\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/make_labeled_video.py:140: FutureWarning: DataFrame.groupby with axis=1 is deprecated. Do `frame.T.groupby(...)` without axis instead.\n", - " Dataframe.groupby(level=\"individuals\", axis=1).size().values // 3\n", - "100%|██████████| 2330/2330 [00:31<00:00, 73.75it/s]\n" - ] - }, - { - "data": { - "text/plain": [ - "[True]" - ] - }, - "execution_count": 12, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "deeplabcut.create_labeled_video(\n", - " config_path,\n", - " [video],\n", - " videotype=\"mp4\",\n", - " shuffle=0,\n", - " color_by=\"individual\",\n", - " keypoints_only=False,\n", - " draw_skeleton=True,\n", - " track_method=\"transformer\"\n", - ")" - ] + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", + " warnings.warn('`layer.apply` is deprecated and '\n", + "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", + " warnings.warn('`layer.apply` is deprecated and '\n", + "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", + " warnings.warn('`layer.apply` is deprecated and '\n", + "/usr/local/lib/python3.11/dist-packages/tensorflow/python/keras/engine/base_layer_v1.py:1694: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.\n", + " warnings.warn('`layer.apply` is deprecated and '\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Activating extracting of PAFs\n", + "Starting to analyze % /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Duration of video [s]: 77.67 , recorded with 30.0 fps!\n", + "Overall # of frames: 2330 found with (before cropping) frame dimensions: 640 480\n", + "Starting to extract posture\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 2330/2330 [01:18<00:00, 29.78it/s]\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "If the tracking is not satisfactory for some videos, consider expanding the training set. You can use the function 'extract_outlier_frames' to extract a few representative outlier frames.\n", + "Epoch 10, train acc: 0.61\n", + "Epoch 10, test acc 0.45\n", + "Epoch 20, train acc: 0.74\n", + "Epoch 20, test acc 0.65\n", + "Epoch 30, train acc: 0.78\n", + "Epoch 30, test acc 0.55\n", + "Epoch 40, train acc: 0.76\n", + "Epoch 40, test acc 0.50\n", + "Epoch 50, train acc: 0.85\n", + "Epoch 50, test acc 0.55\n", + "Epoch 60, train acc: 0.84\n", + "Epoch 60, test acc 0.60\n", + "Epoch 70, train acc: 0.85\n", + "Epoch 70, test acc 0.55\n", + "Epoch 80, train acc: 0.79\n", + "Epoch 80, test acc 0.55\n", + "Epoch 90, train acc: 0.88\n", + "Epoch 90, test acc 0.55\n", + "Epoch 100, train acc: 0.84\n", + "Epoch 100, test acc 0.55\n", + "loading params\n", + "Processing... /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|██████████| 4/4 [00:00<00:00, 483.21it/s]\n", + "/usr/local/lib/python3.11/dist-packages/deeplabcut/refine_training_dataset/stitch.py:934: FutureWarning: Starting with pandas version 3.0 all arguments of to_hdf except for the argument 'path_or_buf' will be keyword-only.\n", + " df.to_hdf(output_name, \"tracks\", format=\"table\", mode=\"w\")\n" + ] + } + ], + "source": [ + "deeplabcut.transformer_reID(\n", + " config_path,\n", + " [video],\n", + " shuffle=0,\n", + " videotype=\"mp4\",\n", + " track_method=\"ellipse\",\n", + " n_triplets=100,\n", + ")" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "uO_yoqN7xiBT" + }, + "source": [ + "now we can make another video with the transformer-guided tracking:\n" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "MBMbRFEMxmi4", + "outputId": "5ca4357a-c8e1-46c6-ecad-141bfce48cc5" + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", + "Plots created! Please check the directory \"plot-poses\" within the video directory\n" + ] } - ], - "metadata": { - "accelerator": "GPU", + ], + "source": [ + "deeplabcut.plot_trajectories(\n", + " config_path,\n", + " [video],\n", + " shuffle=0,\n", + " videotype=\"mp4\",\n", + " track_method=\"transformer\",\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": { "colab": { - "gpuType": "A100", - "include_colab_link": true, - "machine_shape": "hm", - "name": "COLAB_transformer_reID.ipynb", - "provenance": [] + "base_uri": "https://localhost:8080/" }, - "kernelspec": { - "display_name": "Python 3", - "name": "python3" + "id": "vx3e-r1CoXaX", + "outputId": "46cdbd39-d1f6-4b78-abba-7e979740f2a2" + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Starting to process video: /content/demo-me-2021-07-14/videos/videocompressed1.mp4\n", + "Loading /content/demo-me-2021-07-14/videos/videocompressed1.mp4 and data.\n", + "Duration of video [s]: 77.67, recorded with 30.0 fps!\n", + "Overall # of frames: 2330 with cropped frame dimensions: 640 480\n", + "Generating frames and creating video.\n" + ] }, - "language_info": { - "name": "python" + { + "name": "stderr", + "output_type": "stream", + "text": [ + "/usr/local/lib/python3.11/dist-packages/deeplabcut/utils/make_labeled_video.py:140: FutureWarning: DataFrame.groupby with axis=1 is deprecated. Do `frame.T.groupby(...)` without axis instead.\n", + " Dataframe.groupby(level=\"individuals\", axis=1).size().values // 3\n", + "100%|██████████| 2330/2330 [00:31<00:00, 73.75it/s]\n" + ] + }, + { + "data": { + "text/plain": [ + "[True]" + ] + }, + "execution_count": 12, + "metadata": {}, + "output_type": "execute_result" } + ], + "source": [ + "deeplabcut.create_labeled_video(\n", + " config_path,\n", + " [video],\n", + " videotype=\"mp4\",\n", + " shuffle=0,\n", + " color_by=\"individual\",\n", + " keypoints_only=False,\n", + " draw_skeleton=True,\n", + " track_method=\"transformer\"\n", + ")" + ] + } + ], + "metadata": { + "accelerator": "GPU", + "colab": { + "gpuType": "A100", + "include_colab_link": true, + "machine_shape": "hm", + "name": "COLAB_transformer_reID.ipynb", + "provenance": [] + }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2026-02-10", + "last_metadata_updated": "2026-03-06" + }, + "kernelspec": { + "display_name": "Python 3", + "name": "python3" }, - "nbformat": 4, - "nbformat_minor": 0 + "language_info": { + "name": "python" + } + }, + "nbformat": 4, + "nbformat_minor": 0 } diff --git a/examples/JUPYTER/Demo_3D_DeepLabCut.ipynb b/examples/JUPYTER/Demo_3D_DeepLabCut.ipynb index 42242b6664..a92b639131 100644 --- a/examples/JUPYTER/Demo_3D_DeepLabCut.ipynb +++ b/examples/JUPYTER/Demo_3D_DeepLabCut.ipynb @@ -274,6 +274,11 @@ } ], "metadata": { + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-02-28", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/examples/JUPYTER/Demo_labeledexample_MouseReaching.ipynb b/examples/JUPYTER/Demo_labeledexample_MouseReaching.ipynb index 312e0a1505..dfc597b94c 100644 --- a/examples/JUPYTER/Demo_labeledexample_MouseReaching.ipynb +++ b/examples/JUPYTER/Demo_labeledexample_MouseReaching.ipynb @@ -512,6 +512,11 @@ "provenance": [], "version": "0.3.2" }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-02-28", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/examples/JUPYTER/Demo_labeledexample_Openfield.ipynb b/examples/JUPYTER/Demo_labeledexample_Openfield.ipynb index d4458f1111..d3c824b90a 100644 --- a/examples/JUPYTER/Demo_labeledexample_Openfield.ipynb +++ b/examples/JUPYTER/Demo_labeledexample_Openfield.ipynb @@ -445,6 +445,11 @@ "provenance": [], "version": "0.3.2" }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-02-28", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/examples/JUPYTER/Demo_napari.ipynb b/examples/JUPYTER/Demo_napari.ipynb index 6ebbfd7a6c..3a81f75ff1 100644 --- a/examples/JUPYTER/Demo_napari.ipynb +++ b/examples/JUPYTER/Demo_napari.ipynb @@ -529,6 +529,11 @@ "provenance": [], "version": "0.3.2" }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-16", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/examples/JUPYTER/Demo_yourowndata.ipynb b/examples/JUPYTER/Demo_yourowndata.ipynb index 0596204849..d7baf82ca2 100644 --- a/examples/JUPYTER/Demo_yourowndata.ipynb +++ b/examples/JUPYTER/Demo_yourowndata.ipynb @@ -548,6 +548,11 @@ "provenance": [], "version": "0.3.2" }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-16", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/examples/JUPYTER/Docker_TrainNetwork_VideoAnalysis.ipynb b/examples/JUPYTER/Docker_TrainNetwork_VideoAnalysis.ipynb index d6699d623c..80d41613c8 100644 --- a/examples/JUPYTER/Docker_TrainNetwork_VideoAnalysis.ipynb +++ b/examples/JUPYTER/Docker_TrainNetwork_VideoAnalysis.ipynb @@ -331,6 +331,11 @@ "toc_visible": true, "version": "0.3.2" }, + "deeplabcut": { + "ignore": false, + "last_content_updated": "2025-09-16", + "last_metadata_updated": "2026-03-06" + }, "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", diff --git a/pyproject.toml b/pyproject.toml index ea5f250c17..87bb34fa31 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -45,6 +45,7 @@ dependencies = [ "pandas[hdf5,performance]>=2.2,<3", "pillow>=7.1", "pycocotools", + "pydantic>=2,<3", "pyyaml", "ruamel-yaml>=0.15", "scikit-image>=0.17", @@ -110,8 +111,8 @@ Documentation = "https://deeplabcut.github.io/DeepLabCut/README.html" [dependency-groups] dev = [ "coverage", + "nbformat>5", "pre-commit", - "pydantic>=2,<3", "pytest", "pytest-cov", ] diff --git a/tests/tools/docs_and_notebooks_checks/test_check_contracts.py b/tests/tools/docs_and_notebooks_checks/test_check_contracts.py new file mode 100644 index 0000000000..9f4aa2c767 --- /dev/null +++ b/tests/tools/docs_and_notebooks_checks/test_check_contracts.py @@ -0,0 +1,447 @@ +from __future__ import annotations + +import importlib.util +import json +import os +import subprocess +from datetime import date, datetime, timezone +from pathlib import Path +from types import ModuleType + +import pytest + + +# ----------------------------- +# Module loader (tools/ is not necessarily a package) +# ----------------------------- +def load_tool_module() -> ModuleType: + repo_root = Path(__file__).resolve().parents[3] + tool_path = repo_root / "tools" / "docs_and_notebooks_check.py" + assert tool_path.exists(), f"Missing tool: {tool_path}" + + spec = importlib.util.spec_from_file_location("docs_and_notebooks_check", tool_path) + assert spec and spec.loader + mod = importlib.util.module_from_spec(spec) + spec.loader.exec_module(mod) # type: ignore[attr-defined] + return mod + + +@pytest.fixture(scope="session") +def tool() -> ModuleType: + return load_tool_module() + + +# ----------------------------- +# Git helpers for a temp repo +# ----------------------------- +def _run( + cmd: list[str], cwd: Path, env: dict | None = None +) -> subprocess.CompletedProcess: + return subprocess.run( + cmd, cwd=str(cwd), env=env, capture_output=True, text=True, check=True + ) + + +def _git_init(repo: Path) -> None: + _run(["git", "init"], repo) + _run(["git", "config", "user.email", "ci@example.com"], repo) + _run(["git", "config", "user.name", "CI"], repo) + + +def _git_commit(repo: Path, message: str, when_iso: str) -> None: + env = os.environ.copy() + env["GIT_AUTHOR_DATE"] = when_iso + env["GIT_COMMITTER_DATE"] = when_iso + _run(["git", "add", "-A"], repo, env=env) + _run(["git", "commit", "-m", message], repo, env=env) + + +def _write(repo: Path, rel: str, content: str) -> None: + p = repo / rel + p.parent.mkdir(parents=True, exist_ok=True) + p.write_text(content, encoding="utf-8") + + +# ----------------------------- +# Contract tests +# ----------------------------- +def test_marker_constants_exist(tool): + assert hasattr(tool, "META_COMMIT_MARKER") + assert hasattr(tool, "SUGGESTED_TAGGED_COMMIT") + assert tool.META_COMMIT_MARKER in tool.SUGGESTED_TAGGED_COMMIT + + +def test_schema_contract_fields(tool): + # DLCMeta must have new fields and must NOT have old last_git_updated + meta = tool.DLCMeta() + assert hasattr(meta, "last_content_updated") + assert hasattr(meta, "last_metadata_updated") + assert hasattr(meta, "last_verified") + assert hasattr(meta, "verified_for") + assert not hasattr(meta, "last_git_updated") + + +def test_git_content_date_skips_meta_commits(tool, tmp_path: Path): + """ + Contract: last_content_updated is computed from git history excluding metadata commits. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + _write(repo, rel, "# hello\n") + _git_commit(repo, "docs: initial content", "2020-01-01T12:00:00+00:00") + + # meta-only rewrite (simulated) committed with marker + _write( + repo, + rel, + "---\ndeeplabcut:\n last_metadata_updated: 2026-03-01\n---\n# hello\n", + ) + _git_commit( + repo, + f"chore(meta): update {tool.META_COMMIT_MARKER}", + "2026-03-01T12:00:00+00:00", + ) + + # raw touched date = 2026-03-01 + touched = tool.git_last_touched(repo, rel) + assert touched == date(2026, 3, 1) + + # content updated date should skip marker commit => 2020-01-01 + content_date, used_fallback = tool.git_last_content_updated(repo, rel) + assert content_date == date(2020, 1, 1) + assert used_fallback is False + + +def test_git_content_date_fallback_when_only_meta_commits(tool, tmp_path: Path): + """ + If all commits touching the file are meta-marker commits, we fall back to git_last_touched + and flag used_fallback=True. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + _write(repo, rel, "---\ndeeplabcut:\n notes: hi\n---\n") + _git_commit( + repo, + f"chore(meta): init {tool.META_COMMIT_MARKER}", + "2026-03-01T12:00:00+00:00", + ) + + content_date, used_fallback = tool.git_last_content_updated(repo, rel) + assert content_date == date(2026, 3, 1) + assert used_fallback is True + + +def test_scan_is_read_only(tool, tmp_path: Path, monkeypatch): + """ + Contract: report/check (scan_files) must be read-only. + We validate by asserting file content does not change. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + orig = "---\ndeeplabcut:\n last_verified: 2020-01-01\n---\n# hello\n" + _write(repo, rel, orig) + _git_commit(repo, "docs: add page", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + before = (repo / rel).read_text(encoding="utf-8") + records = tool.scan_files(repo, cfg, targets=[rel]) + after = (repo / rel).read_text(encoding="utf-8") + + assert before == after + assert len(records) == 1 + assert records[0].path == rel + assert records[0].kind == "md" + + +def test_update_requires_ack_when_write(tool, tmp_path: Path): + """ + Contract: write mode should refuse unless --ack-meta-commit-marker is provided. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + _write(repo, rel, "# hello\n") + _git_commit(repo, "docs: initial content", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + # should refuse to write without ack + with pytest.raises(SystemExit): + tool.update_files( + repo_root=repo, + cfg=cfg, + targets=[rel], + write=True, + set_content_date_from_git=True, + set_last_verified=None, + set_verified_for=None, + ack_meta_commit_marker=False, + ) + + +def test_update_set_content_date_from_git_only_changes_that_field(tool, tmp_path: Path): + """ + Contract: update --set-content-date-from-git only sets last_content_updated (plus last_metadata_updated when writing), + does NOT override last_verified/verified_for unless explicitly provided. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + initial = ( + "---\n" + "deeplabcut:\n" + " last_verified: 2020-02-02\n" + " verified_for: 3.0.0rc1\n" + "---\n" + "# hello\n" + ) + _write(repo, rel, initial) + _git_commit(repo, "docs: initial content", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + records = tool.update_files( + repo_root=repo, + cfg=cfg, + targets=[rel], + write=True, + set_content_date_from_git=True, + set_last_verified=None, + set_verified_for=None, + ack_meta_commit_marker=True, + ) + assert len(records) == 1 + + # Read back and confirm verified fields unchanged + text = (repo / rel).read_text(encoding="utf-8") + fm, body, _ = tool.read_md_frontmatter(text) + assert isinstance(fm, dict) and tool.DLC_NAMESPACE in fm + meta = fm[tool.DLC_NAMESPACE] + + assert meta["last_verified"] == "2020-02-02" + assert meta["verified_for"] == "3.0.0rc1" + + # last_content_updated should reflect git content date (2020-01-01) + assert meta["last_content_updated"] == "2020-01-01" + + # last_metadata_updated should exist because we wrote + assert "last_metadata_updated" in meta + + +def test_update_set_verified_fields_only_changes_verified(tool, tmp_path: Path): + """ + Contract: update with --set-last-verified / --set-verified-for changes only those fields + (plus last_metadata_updated if writing), and does not set last_content_updated unless requested. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + initial = "---\ndeeplabcut:\n last_content_updated: 2000-01-01\n---\n# hello\n" + _write(repo, rel, initial) + _git_commit(repo, "docs: initial content", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + records = tool.update_files( + repo_root=repo, + cfg=cfg, + targets=[rel], + write=True, + set_content_date_from_git=False, + set_last_verified=date(2026, 3, 5), + set_verified_for="3.0.0rc13", + ack_meta_commit_marker=True, + ) + assert len(records) == 1 + + text = (repo / rel).read_text(encoding="utf-8") + fm, _body, _ = tool.read_md_frontmatter(text) + meta = fm[tool.DLC_NAMESPACE] + + # Verified fields updated + assert meta["last_verified"] == "2026-03-05" + assert meta["verified_for"] == "3.0.0rc13" + + # last_content_updated remains whatever it was (not overwritten) + assert meta["last_content_updated"] == "2000-01-01" + + +def test_normalize_is_explicit_and_marks_would_change(tool, tmp_path: Path): + """ + Contract: normalize is separate and explicit; in dry-run it should mark would_change + if notebook is not already in canonical nbformat output. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/nbs/nb.ipynb" + # Minimal notebook JSON but not in nbformat canonical formatting (indent/newline differences) + raw = ( + "{\n" + ' "cells": [],\n' + ' "metadata": {},\n' + ' "nbformat": 4,\n' + ' "nbformat_minor": 5\n' + "}\n" + ) + _write(repo, rel, raw) + _git_commit(repo, "docs: add notebook", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + # Dry-run normalize: should set would_change True if not normalized + records = tool.normalize_notebooks( + repo_root=repo, + cfg=cfg, + targets=[rel], + write=False, + ack_meta_commit_marker=True, + ) + assert len(records) == 1 + assert records[0].kind == "ipynb" + # may be True depending on canonical formatting differences + assert records[0].would_change + + +def test_write_outputs_contract(tool, tmp_path: Path): + """ + Contract: write_outputs creates both JSON and Markdown files and JSON is schema-valid. + """ + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/page.md" + _write(repo, rel, "# hello\n") + _git_commit(repo, "docs: initial content", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + records = tool.scan_files(repo, cfg, targets=[rel]) + + report = tool.Report( + generated_at=datetime.now(timezone.utc), + repo_root=str(repo), + config_path="in-memory", + totals=tool.summarize(records), + records=records, + ) + + out_dir = tmp_path / "out" + json_path, md_path = tool.write_outputs(report, cfg, out_dir) + + assert json_path.exists() + assert md_path.exists() + + payload = json.loads(json_path.read_text(encoding="utf-8")) + assert payload["schema_version"] == tool.SCHEMA_VERSION + assert "records" in payload and isinstance(payload["records"], list) + assert md_path.read_text(encoding="utf-8").startswith("#") + + +def test_notebook_missing_dlc_namespace_warns_missing_metadata(tool, tmp_path: Path): + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/nbs/nb.ipynb" + # Valid minimal notebook, but no "deeplabcut" namespace under metadata + nb = ( + "{\n" + ' "cells": [],\n' + ' "metadata": {},\n' + ' "nbformat": 4,\n' + ' "nbformat_minor": 5\n' + "}\n" + ) + _write(repo, rel, nb) + _git_commit(repo, "docs: add notebook", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + records = tool.scan_files(repo, cfg, targets=[rel]) + assert len(records) == 1 + r = records[0] + assert r.kind == "ipynb" + assert "missing_metadata" in r.warnings + assert r.meta is None + + +def test_notebook_invalid_dlc_namespace_warns_invalid_metadata(tool, tmp_path: Path): + repo = tmp_path / "repo" + repo.mkdir() + _git_init(repo) + + rel = "docs/nbs/nb.ipynb" + # deeplabcut namespace exists but is invalid: last_verified must be a date + nb = ( + "{\n" + ' "cells": [],\n' + ' "metadata": {\n' + ' "deeplabcut": {\n' + ' "last_verified": "not-a-date"\n' + " }\n" + " },\n" + ' "nbformat": 4,\n' + ' "nbformat_minor": 5\n' + "}\n" + ) + _write(repo, rel, nb) + _git_commit(repo, "docs: add notebook with bad meta", "2020-01-01T12:00:00+00:00") + + cfg = tool.ToolConfig( + version=1, + scan=tool.ScanConfig(include=[rel], exclude=[]), + policy=tool.PolicyConfig(), + ) + + records = tool.scan_files(repo, cfg, targets=[rel]) + assert len(records) == 1 + r = records[0] + assert r.kind == "ipynb" + assert "invalid_metadata" in r.warnings + assert r.meta is None diff --git a/tools/docs_and_notebooks_check.py b/tools/docs_and_notebooks_check.py new file mode 100644 index 0000000000..de8569e478 --- /dev/null +++ b/tools/docs_and_notebooks_check.py @@ -0,0 +1,1111 @@ +"""DeepLabCut docs & notebooks automated checks tool. + +Goals +----- +- SAFE by default: read-only operations in CI (report/check). +- Idempotent updates (update mode) that only touch: + * Notebook-level metadata for .ipynb (never cells/outputs) + * YAML frontmatter for .md docs (optional) +- Uses pydantic schemas with explicit schema_version for validation. +- Aims to be contributor-friendly by default: check mode enforces configured policy, while + surfacing scan/parsing issues without failing unless strict mode is enabled. + +Terminology +----------- +last_content_updated + Computed from git history, excluding metadata-only commits. + (Metadata commits must include META_COMMIT_MARKER in the commit message.) + +last_verified + Human-controlled date indicating the file was verified to work/be accurate. + +verified_for + Human-controlled string, typically the project version (e.g. 3.0.0rc13). + +tier + Optional classification (left unset by default; do not auto-populate). + +Usage modes +----------- +Report (read-only): + python tools/docs_and_notebooks_check.py report + +Check (read-only; policy enforcement): + python tools/docs_and_notebooks_check.py check + + Runs scans and evaluates configured policy rules. + Exits non-zero for policy violations. + Scan/parsing errors are always reported in console / JSON / Markdown output, + but are non-fatal by default unless strict mode is enabled or they imply a + policy violation. + +Update content-date field from git (write mode; requires --write): + python tools/docs_and_notebooks_check.py update --write --set-content-date-from-git + +Update verification fields for selected targets (write mode): + python tools/docs_and_notebooks_check.py update --write --targets docs/page.md \ + --set-last-verified today --set-verified-for 3.0.0rc13 + +Normalize notebooks deterministically (explicit churn; write mode): + python tools/docs_and_notebooks_check.py normalize --write --targets docs/notebook.ipynb + + +Configuration +------------- +Uses tools/docs_and_notebooks_report_config.yml by default. + +Outputs +------- +- docs_nb_checks.json: machine-readable report +- docs_nb_checks.md: human-readable summary + +Notes for CI +------------ +- Ensure actions/checkout uses fetch-depth: 0 (or sufficiently deep), + otherwise git log may not see history. +- Requires: + - pydantic>=2,<3 + - PyYAML + - nbformat>=5 + to be installed in the environment. + Recommended : install in CI job directly (pip install pydantic pyyaml nbformat) rather than adding to requirements, since these are only needed for this tool. +""" +# tools/docs_and_notebooks_check.py +from __future__ import annotations + +import argparse +import fnmatch +import json +import os +import re +import subprocess +from datetime import date, datetime, timezone +from pathlib import Path +from typing import Any, Dict, List, Literal, Optional, Sequence, Tuple + +try: + import yaml # PyYAML +except Exception: + yaml = None + +try: + from pydantic import BaseModel, ConfigDict, Field, ValidationError +except Exception: # pragma: no cover + raise RuntimeError("Pydantic is required to run this script") +try: + import nbformat + from nbformat.validator import NotebookValidationError +except Exception: + raise RuntimeError("nbformat is required to read/write .ipynb files") + +SCHEMA_VERSION = 1 +DLC_NAMESPACE = "deeplabcut" +OUTPUT_FILENAME = "docs_nb_checks" +SCRIPT_DIR = Path(__file__).resolve().parent +DEFAULT_CFG = SCRIPT_DIR / "docs_and_notebooks_report_config.yml" + + +# ----------------------------- +# Metadata commit marker / guidance +# ----------------------------- +# IMPORTANT: +# Metadata-only updates and notebook normalization rewrite files and will change +# "git last touched" timestamps. To preserve meaningful "content age", all such +# commits must include this marker in the commit message. +META_COMMIT_MARKER = "chore(metadata)" +SUGGESTED_TAGGED_COMMIT = f"{META_COMMIT_MARKER}: update docs/notebooks metadata" + + +# ----------------------------- +# Pydantic schemas +# ----------------------------- + + +class DLCMeta(BaseModel): + """Metadata embedded in files under the `deeplabcut` namespace.""" + + model_config = ConfigDict(extra="allow") + + # Tool-managed: last meaningful content update date (excluding metadata commits) + last_content_updated: Optional[date] = None + + # Optional tool-managed: last time metadata/normalization was performed + last_metadata_updated: Optional[date] = None + # Optional human-managed verification fields + last_verified: Optional[date] = None + # Version or other string indicating what this file was verified for (e.g. "3.0.0rc13") + verified_for: Optional[str] = None + # Extra metadata fields for later usage (e.g. allowlist tier classification), but not currently used by the tool + tier: Optional[str] = None + ignore: bool = False + notes: Optional[str] = None + + +class ScanConfig(BaseModel): + include: List[str] = Field(default_factory=list) + exclude: List[str] = Field(default_factory=list) + + +class PolicyConfig(BaseModel): + warn_if_content_older_than_days: int = 365 + warn_if_verified_older_than_days: int = 365 + missing_last_verified_is_warning: bool = True + + # Strict-mode toggle: if true, scan/parsing errors also fail `check` + fail_on_scan_errors: bool = False + + # Allowlists for strict checks (start empty; ratchet later) + require_metadata: List[str] = Field(default_factory=list) + require_recent_verification: List[str] = Field(default_factory=list) + + require_notebook_normalized: List[str] = Field(default_factory=list) + + +class ToolConfig(BaseModel): + version: int = 1 + scan: ScanConfig + policy: PolicyConfig + + +class FileRecord(BaseModel): + path: str + kind: str # ipynb | md | other + + # Computed from git (excluding metadata-only commits) + last_content_updated: Optional[date] = None + # Debug-only: raw git last touched (may be metadata commit) + last_git_touched: Optional[date] = None + + # Read from file metadata/frontmatter + meta: Optional[DLCMeta] = None + + # Derived + days_since_content_update: Optional[int] = None + days_since_verified: Optional[int] = None + + warnings: List[str] = Field(default_factory=list) + errors: List[str] = Field(default_factory=list) + + # If update mode would change file + would_change: bool = False + + +class Report(BaseModel): + schema_version: int = SCHEMA_VERSION + generated_at: datetime + repo_root: str + config_path: str + + totals: Dict[str, int] + records: List[FileRecord] + + +# Rebuild models due to __future__ annotations +DLCMeta.model_rebuild() +ScanConfig.model_rebuild() +PolicyConfig.model_rebuild() +ToolConfig.model_rebuild() +FileRecord.model_rebuild() +Report.model_rebuild() +# ----------------------------- +# Helpers +# ----------------------------- + + +def _iso_today() -> date: + return datetime.now(timezone.utc).date() + + +def _run_git(args: Sequence[str], cwd: Path) -> Tuple[int, str, str]: + p = subprocess.run( + ["git", *args], + cwd=str(cwd), + stdout=subprocess.PIPE, + stderr=subprocess.PIPE, + text=True, + ) + return p.returncode, p.stdout.strip(), p.stderr.strip() + + +def find_repo_root(start: Path) -> Path: + cur = start.resolve() + for _ in range(50): + if (cur / ".git").exists(): + return cur + if cur.parent == cur: + break + cur = cur.parent + code, out, _err = _run_git(["rev-parse", "--show-toplevel"], cwd=start) + if code == 0 and out: + return Path(out).resolve() + raise RuntimeError("Could not locate repository root") + + +def glob_paths(repo_root: Path, patterns: List[str]) -> List[Path]: + results: List[Path] = [] + for pat in patterns: + results.extend(repo_root.glob(pat)) + return sorted({p.resolve() for p in results if p.is_file()}) + + +def is_excluded(rel_path: str, exclude_patterns: List[str]) -> bool: + return any(fnmatch.fnmatch(rel_path, pat) for pat in exclude_patterns) + + +def file_kind(path: Path) -> str: + s = path.suffix.lower() + if s == ".ipynb": + return "ipynb" + if s in {".md", ".markdown"}: + return "md" + return "other" + + +def _parse_git_iso_date(out: str) -> Optional[date]: + out = (out or "").strip() + if not out: + return None + + try: + return date.fromisoformat(out) + except Exception: + pass + + try: + if out.endswith("Z"): + out = out[:-1] + "+00:00" + return datetime.fromisoformat(out).date() + except Exception: + return None + + +def _git_log_date( + repo_root: Path, rel_path: str, extra_args: Sequence[str] = () +) -> Optional[date]: + args = [ + "log", + "-1", + "--date=short", + "--format=%cd", + *extra_args, + "--", + rel_path, + ] + code, out, _err = _run_git(args, cwd=repo_root) + if code != 0: + return None + return _parse_git_iso_date(out) + + +def git_last_touched(repo_root: Path, rel_path: str) -> Optional[date]: + return _git_log_date(repo_root, rel_path) + + +def git_last_content_updated( + repo_root: Path, rel_path: str +) -> Tuple[Optional[date], bool]: + d = _git_log_date( + repo_root, + rel_path, + extra_args=[ + "--fixed-strings", + "--invert-grep", + "--grep", + META_COMMIT_MARKER, + ], + ) + if d is not None: + return d, False + return git_last_touched(repo_root, rel_path), True + + +FRONTMATTER_RE = re.compile(r"^---\s*$") + + +def read_md_frontmatter(text: str) -> Tuple[Optional[dict], str, Optional[str]]: + lines = text.splitlines(keepends=True) + if not lines or not FRONTMATTER_RE.match(lines[0]): + return None, text, None + + end_idx = None + for i in range(1, min(len(lines), 5000)): + if FRONTMATTER_RE.match(lines[i]): + end_idx = i + break + + if end_idx is None: + return None, text, "unterminated_markdown_frontmatter" + + fm_text = "".join(lines[1:end_idx]) + body = "".join(lines[end_idx + 1 :]) + + if yaml is None: + raise RuntimeError("PyYAML is required to parse Markdown frontmatter") + + fm = yaml.safe_load(fm_text) if fm_text.strip() else {} + if not isinstance(fm, dict): + return None, text, "markdown_frontmatter_not_mapping" + + return fm, body, None + + +def dump_md_frontmatter(frontmatter: dict, body: str) -> str: + if yaml is None: + raise RuntimeError("PyYAML is required to write Markdown frontmatter") + fm_text = yaml.safe_dump(frontmatter, sort_keys=False, allow_unicode=True) + body_to_write = body + if body_to_write.startswith("\n"): + body_to_write = body_to_write[1:] + return "---\n" + fm_text + "---\n" + body_to_write + + +def read_ipynb_meta(path: Path) -> tuple[Any, dict, bool]: + """ + Read a notebook using nbformat. + Returns (notebook_node, deeplabcut_meta_dict, has_dlc_namespace). + """ + nb = nbformat.read(str(path), as_version=4) + + meta = getattr(nb, "metadata", {}) or {} + has_dlc = DLC_NAMESPACE in meta + + raw_dlc_meta = meta.get(DLC_NAMESPACE) + return nb, raw_dlc_meta, has_dlc + + +def notebook_is_normalized(path: Path, nb: Any) -> bool: + original = path.read_text(encoding="utf-8") + # Normalize newline style so CRLF vs LF differences do not cause false mismatches + original_normalized = original.replace("\r\n", "\n").replace("\r", "\n") + normalized = nbformat.writes(nb, version=4, indent=2, ensure_ascii=False) + "\n" + return original_normalized == normalized + + +def write_ipynb_meta(path: Path, nb: Any) -> None: + """ + Write a notebook using nbformat. + + Note: nbformat writes JSON in a canonical form; it *will* rewrite the file, + so expect diffs if the notebook wasn't previously normalized to the same style. + """ + # Validate before writing (optional but recommended) + nbformat.validate(nb) + + # Use a stable indentation to reduce churn (choose 2 if your repo tends that way) + text = nbformat.writes(nb, version=4, indent=2, ensure_ascii=False) + + path.write_text(text + "\n", encoding="utf-8") + + +def parse_dlc_meta(raw: Any) -> tuple[Optional[DLCMeta], bool]: + # returns (meta, valid) + if raw is None or not isinstance(raw, dict): + return None, False + try: + return DLCMeta.model_validate(raw), True + except ValidationError: + return None, False + + +def meta_to_jsonable(meta: DLCMeta) -> dict: + """ + Return JSON-serializable metadata (dates become ISO strings). + This prevents json.dumps() from failing when writing .ipynb files. + """ + return meta.model_dump(mode="json", exclude_none=True) + + +def compute_days_since(d: Optional[date], today: date) -> Optional[int]: + return None if d is None else (today - d).days + + +def match_allowlist(rel_path: str, allowlist: List[str]) -> bool: + # Support exact matches or glob patterns + return any(pat == rel_path or fnmatch.fnmatch(rel_path, pat) for pat in allowlist) + + +# ----------------------------- +# Core scanning +# ----------------------------- + + +def load_config(config_path: Path) -> ToolConfig: + if yaml is None: + raise RuntimeError("PyYAML is required (pip install pyyaml)") + raw = yaml.safe_load(config_path.read_text(encoding="utf-8")) + return ToolConfig.model_validate(raw) + + +def scan_files( + repo_root: Path, cfg: ToolConfig, targets: Optional[List[str]] = None +) -> List[FileRecord]: + today = _iso_today() + paths = glob_paths(repo_root, cfg.scan.include) + records: List[FileRecord] = [] + target_set = None + if targets: + target_set = set(t.replace(os.sep, "/") for t in targets) + + for p in paths: + rel = str(p.resolve().relative_to(repo_root)).replace(os.sep, "/") + if is_excluded(rel, cfg.scan.exclude): + continue + if target_set is not None and rel not in target_set: + continue + kind = file_kind(p) + rec = FileRecord(path=rel, kind=kind) + + rec.last_git_touched = git_last_touched(repo_root, rel) + rec.last_content_updated, used_fallback = git_last_content_updated( + repo_root, rel + ) + rec.days_since_content_update = compute_days_since( + rec.last_content_updated, today + ) + if used_fallback: + rec.warnings.append("content_date_fallback_to_git_touched") + + try: + if kind == "ipynb": + nb, raw_meta, has_dlc = read_ipynb_meta(p) + + try: + nbformat.validate(nb) + except NotebookValidationError as e: + rec.errors.append(f"nbformat_invalid: {e}") + + try: + if not notebook_is_normalized(p, nb): + rec.warnings.append("notebook_not_normalized") + except Exception as e: + rec.errors.append(f"notebook_normalization_check_failed: {e}") + + if not has_dlc: + rec.meta = None + rec.warnings.append("missing_metadata") + else: + rec.meta, valid = parse_dlc_meta(raw_meta) + if not valid: + rec.meta = None + rec.warnings.append("invalid_metadata") + + elif kind == "md": + text = p.read_text(encoding="utf-8") + fm, _body, fm_error = read_md_frontmatter(text) + + if fm_error: + rec.meta = None + rec.warnings.append("invalid_metadata") + rec.errors.append(f"markdown_frontmatter_invalid: {fm_error}") + else: + fm = fm or {} + has_dlc = DLC_NAMESPACE in fm + raw = fm.get(DLC_NAMESPACE) + + if not has_dlc: + rec.meta = None + rec.warnings.append("missing_metadata") + else: + rec.meta, valid = parse_dlc_meta(raw) + if not valid: + rec.meta = None + rec.warnings.append("invalid_metadata") + + else: + rec.meta = None + + except Exception as e: + rec.errors.append(f"metadata_read_failed: {e}") + + # ignore=True means: keep reporting diagnostics, but skip freshness/policy logic + if rec.meta and rec.meta.ignore: + records.append(rec) + continue + + last_verified = rec.meta.last_verified if rec.meta else None + rec.days_since_verified = compute_days_since(last_verified, today) + + # Future dates are data errors + if rec.last_content_updated is not None and rec.last_content_updated > today: + rec.errors.append("future_last_content_updated") + if rec.meta and rec.meta.last_metadata_updated is not None: + if rec.meta.last_metadata_updated > today: + rec.errors.append("future_last_metadata_updated") + if last_verified is not None and last_verified > today: + rec.errors.append("future_last_verified") + + pol = cfg.policy + + if ( + rec.days_since_content_update is not None + and rec.days_since_content_update > pol.warn_if_content_older_than_days + ): + rec.warnings.append(f"content_stale>{pol.warn_if_content_older_than_days}d") + + if last_verified is None and pol.missing_last_verified_is_warning: + rec.warnings.append("missing_last_verified") + elif ( + rec.days_since_verified is not None + and rec.days_since_verified > pol.warn_if_verified_older_than_days + ): + rec.warnings.append( + f"verified_stale>{pol.warn_if_verified_older_than_days}d" + ) + + records.append(rec) + + return records + + +# ----------------------------- +# Update mode +# ----------------------------- +def _require_meta_marker_ack(write: bool, ack_marker: bool) -> None: + """ + Guardrail: writing metadata/normalization without the marker convention will + destroy the meaning of content freshness signals. Require an explicit ack. + """ + if not write: + return + if ack_marker: + return + raise SystemExit( + "Refusing to write without acknowledging metadata-commit convention.\n" + "Re-run with --ack-meta-commit-marker and commit with:\n" + f" {SUGGESTED_TAGGED_COMMIT}\n" + ) + + +def update_files( + repo_root: Path, + cfg: ToolConfig, + targets: Optional[List[str]], + write: bool, + set_content_date_from_git: bool, + set_last_verified: Optional[date], + set_verified_for: Optional[str], + ack_meta_commit_marker: bool, +) -> List[FileRecord]: + today = _iso_today() + records = scan_files(repo_root, cfg, targets=targets) + target_set = set(t.replace(os.sep, "/") for t in targets) if targets else None + + for rec in records: + if rec.kind not in {"ipynb", "md"}: + continue + if rec.meta and rec.meta.ignore: + continue + if target_set is not None and rec.path not in target_set: + continue + + meta = rec.meta or DLCMeta() + + # Build the desired metadata WITHOUT touching last_metadata_updated. + if set_content_date_from_git and rec.last_content_updated is not None: + meta.last_content_updated = rec.last_content_updated + + if set_last_verified is not None: + meta.last_verified = set_last_verified + if set_verified_for is not None: + meta.verified_for = set_verified_for + + desired_base = meta_to_jsonable(meta) + abs_path = repo_root / rec.path + changed = False + + if rec.kind == "ipynb": + nb, _raw, _has_dlc = read_ipynb_meta(abs_path) + nb_meta = nb.setdefault("metadata", {}) + prev = nb_meta.get(DLC_NAMESPACE, {}) + if not isinstance(prev, dict): + prev = {} + + merged_base = dict(prev) + merged_base.update(desired_base) + + if merged_base != prev: + changed = True + if write: + _require_meta_marker_ack( + write=True, ack_marker=ack_meta_commit_marker + ) + + meta.last_metadata_updated = today + desired_final = meta_to_jsonable(meta) + + merged_final = dict(prev) + merged_final.update(desired_final) + nb_meta[DLC_NAMESPACE] = merged_final + write_ipynb_meta(abs_path, nb) + + elif rec.kind == "md": + text = abs_path.read_text(encoding="utf-8") + fm, body, fm_error = read_md_frontmatter(text) + if fm_error: + msg = f"markdown_frontmatter_invalid: {fm_error}" + if msg not in rec.errors: + rec.errors.append(msg) + continue + + fm = fm or {} + + prev = fm.get(DLC_NAMESPACE, {}) + if not isinstance(prev, dict): + prev = {} + + merged_base = dict(prev) + merged_base.update(desired_base) + + if merged_base != prev: + changed = True + if write: + _require_meta_marker_ack( + write=True, ack_marker=ack_meta_commit_marker + ) + + meta.last_metadata_updated = today + desired_final = meta_to_jsonable(meta) + + merged_final = dict(prev) + merged_final.update(desired_final) + fm[DLC_NAMESPACE] = merged_final + abs_path.write_text(dump_md_frontmatter(fm, body), encoding="utf-8") + + rec.would_change = changed + rec.meta = meta + rec.days_since_verified = compute_days_since(meta.last_verified, today) + + return records + + +# ----------------------------- +# Notebook formatting +# ----------------------------- +def normalize_notebooks( + repo_root: Path, + cfg: ToolConfig, + targets: Optional[List[str]], + write: bool, + ack_meta_commit_marker: bool, +) -> List[FileRecord]: + """ + Normalize notebooks deterministically (canonical nbformat JSON). + This is intentionally separated from update() because it causes churn. + """ + _require_meta_marker_ack(write=write, ack_marker=ack_meta_commit_marker) + records = scan_files(repo_root, cfg, targets=targets) + today = _iso_today() + + for rec in records: + if rec.kind != "ipynb": + continue + if rec.meta and rec.meta.ignore: + continue + + abs_path = repo_root / rec.path + try: + nb, _raw, _has_dlc = read_ipynb_meta(abs_path) + nbformat.validate(nb) + + if not notebook_is_normalized(abs_path, nb): + rec.would_change = True + if write: + # Update embedded maintenance timestamp + meta = rec.meta or DLCMeta() + meta.last_metadata_updated = today + + nb_meta = nb.setdefault("metadata", {}) + prev = nb_meta.get(DLC_NAMESPACE, {}) + if not isinstance(prev, dict): + prev = {} + merged = dict(prev) + merged.update(meta_to_jsonable(meta)) + nb_meta[DLC_NAMESPACE] = merged + + # Write to persist metadata update (still canonical) + write_ipynb_meta(abs_path, nb) + rec.meta = meta + + except Exception as e: + rec.errors.append(f"normalize_failed: {e}") + + return records + + +# ----------------------------- +# Output formatting +# ----------------------------- + + +def summarize(records: List[FileRecord]) -> Dict[str, int]: + return { + "files": len(records), + "warnings": sum(1 for r in records if r.warnings), + "errors": sum(1 for r in records if r.errors), + "missing_metadata": sum(1 for r in records if "missing_metadata" in r.warnings), + "missing_last_verified": sum( + 1 for r in records if "missing_last_verified" in r.warnings + ), + "content_stale": sum( + 1 for r in records if any(w.startswith("content_stale") for w in r.warnings) + ), + "verified_stale": sum( + 1 + for r in records + if any(w.startswith("verified_stale") for w in r.warnings) + ), + } + + +def to_markdown(report: Report, cfg: ToolConfig) -> str: + pol = cfg.policy + t = report.totals + lines: List[str] = [] + + lines.append("# 🌡️ DeepLabCut freshness report\n") + lines.append(f"Generated: {report.generated_at.isoformat()}\n") + lines.append(f"Schema: v{report.schema_version}\n\n") + + lines.append("## Summary\n") + lines.append(f"- Files scanned: **{t['files']}**\n") + lines.append(f"- Files with warnings: **{t['warnings']}**\n") + lines.append(f"- Files with scanning errors: **{t['errors']}**\n") + lines.append(f"- Missing metadata: **{t['missing_metadata']}**\n") + lines.append(f"- Missing last_verified: **{t['missing_last_verified']}**\n") + lines.append( + f"- Content-stale (> {pol.warn_if_content_older_than_days}d): **{t['content_stale']}**\n" + ) + lines.append( + f"- Verification-stale (> {pol.warn_if_verified_older_than_days}d): **{t['verified_stale']}**\n\n" + ) + + def fmt_date(d: Optional[date]) -> str: + return d.isoformat() if d else "-" + + warn_recs = [ + r for r in report.records if r.warnings and not (r.meta and r.meta.ignore) + ] + warn_recs.sort( + key=lambda r: ( + -(r.days_since_verified or -1), + -(r.days_since_content_update or -1), + r.path, + ) + ) + + if warn_recs: + lines.append("## Warnings\n") + for r in warn_recs: + meta = r.meta + lines.append(f"- **{r.path}** ({r.kind})\n") + lines.append( + f" - last_content_updated: {fmt_date(r.last_content_updated)} " + f"(days: {r.days_since_content_update if r.days_since_content_update is not None else '-'})\n" + ) + if r.last_git_touched: + lines.append(f" - last_git_touched: {fmt_date(r.last_git_touched)}\n") + if meta and meta.last_metadata_updated: + lines.append( + f" - last_metadata_updated: {fmt_date(meta.last_metadata_updated)}\n" + ) + lv = meta.last_verified if meta else None + lines.append( + f" - last_verified: {fmt_date(lv)} " + f"(days: {r.days_since_verified if r.days_since_verified is not None else '-'})\n" + ) + if meta and meta.verified_for: + lines.append(f" - verified_for: {meta.verified_for}\n") + if meta and meta.tier: + lines.append(f" - tier: {meta.tier}\n") + lines.append(f" - warnings: {', '.join(r.warnings)}\n") + if r.errors: + lines.append(f" - errors: {', '.join(r.errors)}\n") + lines.append("\n") + + err_recs = [r for r in report.records if r.errors] + if err_recs: + lines.append("## Scan errors\n") + for r in err_recs: + lines.append(f"- **{r.path}**: {', '.join(r.errors)}\n") + lines.append("\n") + + lines.append("## Notes\n") + lines.append( + "- 'Out of date' does not necessarily mean 'broken'. Use this as a triage signal.\n" + ) + lines.append( + "- last_git_touched / last_content_updated are computed from git history. last_verified is human-controlled.\n\n" + ) + lines.append( + "- In `check` mode, scan/parsing errors are reported for visibility but do not " + "fail by default unless strict mode is enabled or they trigger an enforced policy rule.\n" + ) + return "".join(lines) + + +def write_outputs(report: Report, cfg: ToolConfig, out_dir: Path) -> Tuple[Path, Path]: + out_dir.mkdir(parents=True, exist_ok=True) + json_path = out_dir / f"{OUTPUT_FILENAME}.json" + md_path = out_dir / f"{OUTPUT_FILENAME}.md" + + payload = report.model_dump(mode="json") + + json_path.write_text( + json.dumps(payload, indent=2, ensure_ascii=False) + "\n", encoding="utf-8" + ) + md_path.write_text(to_markdown(report, cfg), encoding="utf-8") + return json_path, md_path + + +# ----------------------------- +# Check enforcement +# ----------------------------- +def enforce(cfg: ToolConfig, records: List[FileRecord]) -> List[str]: + pol = cfg.policy + violations: List[str] = [] + today = _iso_today() + + for r in records: + if r.meta and r.meta.ignore: + continue + if r.kind not in {"ipynb", "md"}: + continue + + has_invalid_metadata = "invalid_metadata" in (r.warnings or []) + + if match_allowlist(r.path, pol.require_metadata): + if has_invalid_metadata: + violations.append(f"{r.path}: invalid metadata") + elif r.meta is None: + violations.append(f"{r.path}: missing metadata") + + if match_allowlist(r.path, pol.require_recent_verification): + if has_invalid_metadata: + violations.append(f"{r.path}: invalid metadata") + else: + lv = r.meta.last_verified if r.meta else None + if lv is None: + violations.append(f"{r.path}: missing last_verified") + else: + days = (today - lv).days + if days > pol.warn_if_verified_older_than_days: + violations.append( + f"{r.path}: last_verified is {days}d old " + f"(> {pol.warn_if_verified_older_than_days}d)" + ) + + if r.kind == "ipynb" and match_allowlist( + r.path, pol.require_notebook_normalized + ): + if "notebook_not_normalized" in (r.warnings or []): + violations.append( + f"{r.path}: notebook is not normalized (run update/format)" + ) + + return violations + + +# ----------------------------- +# CLI +# ----------------------------- + + +def parse_date_token(token: str) -> date: + token = token.strip().lower() + if token in {"today", "now"}: + return _iso_today() + return date.fromisoformat(token) + + +def collect_scan_issues( + records: List[FileRecord], target: Literal["errors", "warnings"] +) -> List[str]: + items: List[str] = [] + for r in records: + for e in getattr(r, target, []): + items.append(f"{r.path}: {e}") + return items + + +def main(argv: Optional[Sequence[str]] = None) -> int: + parser = argparse.ArgumentParser( + description="DeepLabCut checks tool (docs + notebooks)" + ) + parser.add_argument( + "--config", default=str(DEFAULT_CFG), help="Path to YAML config file" + ) + parser.add_argument( + "--no-step-summary", + action="store_true", + help="Do not write to GITHUB_STEP_SUMMARY", + ) + parser.add_argument( + "--out-dir", default=f"tmp/{OUTPUT_FILENAME}", help="Directory to write outputs" + ) + + sub = parser.add_subparsers(dest="cmd", required=True) + rep = sub.add_parser("report", help="Generate staleness report (read-only)") + rep.add_argument( + "--targets", + nargs="*", + help="Optional list of relative file paths to scan (limits scan to these files)", + ) + + chk = sub.add_parser( + "check", + help=( + "Run scans + policy checks (read-only). " + "Fails on enforced policy violations; scan errors are non-fatal by default." + ), + ) + chk.add_argument( + "--targets", + nargs="*", + help="Optional list of relative file paths to scan (limits scan to these files)", + ) + chk.add_argument( + "--strict-mode", + action="store_true", + help="Enable failure on scan/parsing errors (overrides config for this run)", + ) + + up = sub.add_parser( + "update", help="Update metadata/frontmatter (write mode requires --write)" + ) + up.add_argument( + "--write", + action="store_true", + help="Actually write changes (otherwise dry-run)", + ) + up.add_argument( + "--set-content-date-from-git", + action="store_true", + help="Set embedded last_content_updated from computed git content date", + ) + up.add_argument( + "--targets", nargs="*", help="Optional list of relative file paths to update" + ) + up.add_argument("--set-last-verified", default=None, help="YYYY-MM-DD or 'today'") + up.add_argument("--set-verified-for", default=None, help="String like 3.0.0rc13") + up.add_argument( + "--ack-meta-commit-marker", + action="store_true", + help=f"Acknowledge that you will commit changes using marker: {META_COMMIT_MARKER}", + ) + + norm = sub.add_parser( + "normalize", + help="Normalize notebooks deterministically (write mode requires --write)", + ) + norm.add_argument( + "--write", + action="store_true", + help="Actually write changes (otherwise dry-run)", + ) + norm.add_argument( + "--targets", + nargs="*", + help="Optional list of relative notebook paths to normalize", + ) + norm.add_argument( + "--ack-meta-commit-marker", + action="store_true", + help=f"Acknowledge that you will commit changes using marker: {META_COMMIT_MARKER}", + ) + + args = parser.parse_args(list(argv) if argv is not None else None) + + config_path = Path(args.config) + repo_root = find_repo_root(Path.cwd()) + cfg = load_config(config_path) + out_dir = Path(args.out_dir) + + if args.cmd in {"report", "check"}: + records = scan_files(repo_root, cfg, targets=getattr(args, "targets", None)) + elif args.cmd == "update": + lv = ( + parse_date_token(args.set_last_verified) if args.set_last_verified else None + ) + records = update_files( + repo_root, + cfg, + targets=args.targets, + write=bool(args.write), + set_content_date_from_git=bool(args.set_content_date_from_git), + set_last_verified=lv, + set_verified_for=args.set_verified_for, + ack_meta_commit_marker=bool(args.ack_meta_commit_marker), + ) + if args.write: + print(f"\nSuggested commit message:\n {SUGGESTED_TAGGED_COMMIT}\n") + + else: # normalize + records = normalize_notebooks( + repo_root, + cfg, + targets=args.targets, + write=bool(args.write), + ack_meta_commit_marker=bool(args.ack_meta_commit_marker), + ) + if args.write: + print(f"\nSuggested commit message:\n {SUGGESTED_TAGGED_COMMIT}\n") + + report = Report( + generated_at=datetime.now(timezone.utc), + repo_root=str(repo_root), + config_path=str(config_path), + totals=summarize(records), + records=records, + ) + + json_path, md_path = write_outputs(report, cfg, out_dir) + + # Emit GitHub Actions job summary if available + emit_summary = not getattr(args, "no_step_summary", False) + step_summary = os.environ.get("GITHUB_STEP_SUMMARY") + if emit_summary and step_summary and md_path.exists(): + try: + content = md_path.read_text(encoding="utf-8") + # snippet = "\n".join(content.splitlines()[:220]) + "\n" + snippet = "\n".join(content.splitlines()[:]) + "\n" + Path(step_summary).write_text(snippet, encoding="utf-8") + except Exception: + pass + + scan_errors = collect_scan_issues(records, target="errors") + if scan_errors: + print("\nScan errors detected (non-fatal by default):") + for item in scan_errors[:20]: + print(f"- {item}") + if len(scan_errors) > 20: + print(f"... and {len(scan_errors) - 20} more (see report for full details)") + + if args.cmd == "check": + violations = enforce(cfg, records) + if violations: + print("Policy violations:") + for v in violations: + print(f"- {v}") + return 2 + strict_mode = bool((args.strict_mode) or cfg.policy.fail_on_scan_errors) + if strict_mode and scan_errors: + print("Strict mode enabled: failing due to scan/parsing errors.") + return 1 + + # Non-zero if metadata parsing errors occurred for non-report/check commands + if args.cmd not in {"report", "check"} and any(r.errors for r in records): + return 1 + else: + print(f"\nReport generated:") + print(f"- JSON: {json_path}") + print(f"- Markdown: {md_path}") + + if any(r.warnings for r in records): + print("Warnings detected; see report for details.") + if any(r.errors for r in records): + print("Scan errors detected; see report for details.") + return 0 + + +if __name__ == "__main__": + raise SystemExit(main()) diff --git a/tools/docs_and_notebooks_report_config.yml b/tools/docs_and_notebooks_report_config.yml new file mode 100644 index 0000000000..5342f4dfb9 --- /dev/null +++ b/tools/docs_and_notebooks_report_config.yml @@ -0,0 +1,26 @@ +version: 1 + +scan: + include: + - "examples/COLAB/**/*.ipynb" + - "examples/JUPYTER/**/*.ipynb" + - "docs/**/*.md" + - "docs/**/*.ipynb" # if notebooks get added to docs (Jupyter Book supports this) + exclude: + - "**/.ipynb_checkpoints/**" + - "**/_build/**" + - "**/build/**" + +policy: + warn_if_content_older_than_days: 365 + warn_if_verified_older_than_days: 365 + missing_last_verified_is_warning: true + + # Ratchet lists for tiered verification requirements. + # Tiers have to be determined, and crucial targets identified. + # Then specific policies can be set for each tier, + # e.g. requiring more recent verification for higher tiers, + # or requiring verification for more recent versions. + require_metadata: [] + require_recent_verification: [] + require_notebook_normalized: [] diff --git a/tools/docs_and_notebooks_tool_README.md b/tools/docs_and_notebooks_tool_README.md new file mode 100644 index 0000000000..fa27c5dc8f --- /dev/null +++ b/tools/docs_and_notebooks_tool_README.md @@ -0,0 +1,190 @@ +# Docs & Notebooks Checks Tool + +This tool scans DeepLabCut documentation pages and notebooks and produces **two independent signals**: + +- **`last_content_updated`**: computed from git history as the last *meaningful content* update **excluding metadata-only commits**. +- **`last_verified`**: a human-controlled date indicating the content was verified to work/be accurate. + +In addition, the tool can optionally track: + +- **`last_metadata_updated`**: when the tool last performed a metadata/normalization write (helps explain “file changed” without implying content changed). +- **`verified_for`**: a human-controlled string indicating what the content was verified against (e.g. `3.0.0rc13`). + +The tool is designed to be: + +- **Safe by default**: CI should run **read-only** modes (`report` / `check`). +- **Deterministic**: stable outputs and normalized notebook formatting when explicitly requested. +- **Future-proof**: versioned Pydantic schemas (`schema_version`). + +--- + +## What gets scanned + +Default include patterns are defined in `tools/docs_and_notebooks_report_config.yml`. +Typical patterns include: + +- `examples/COLAB/**/*.ipynb` +- `examples/JUPYTER/**/*.ipynb` +- `docs/**/*.md` +- `docs/**/*.ipynb` (if notebooks are added under docs) + +You can further restrict the scan via `--targets`. + +--- + +## Metadata storage locations + +### Notebooks (`.ipynb`) + +The tool **only** reads/writes **top-level notebook metadata** under the `deeplabcut` namespace. + +> [!IMPORTANT] +> It never edits notebook cells, outputs, or execution counts. + +Example (excerpt): + +```json +{ + "metadata": { + "deeplabcut": { + "last_content_updated": "2020-01-01", + "last_metadata_updated": "2026-03-05", + "last_verified": "2026-02-20", + "verified_for": "3.0.0rc13", + "ignore": false + } + } +} +``` + +> [!NOTE] +> `tier` is intentionally optional and is not auto-populated. + +### Markdown (`.md`) + +The tool reads/writes YAML frontmatter at the top of the file (if present): + +```yaml +--- +deeplabcut: + last_content_updated: 2020-01-01 + last_metadata_updated: 2026-03-05 + last_verified: 2026-02-20 + verified_for: 3.0.0rc13 + ignore: false +--- +``` + +If a doc page has **no** frontmatter, the tool can still report staleness (read-only), and `update` can add/modify metadata when explicitly requested. + +--- + +## The metadata-commit marker (critical) + +Because metadata updates and notebook normalization can rewrite files, they would normally make git (correctly) report that the file was “updated now”. + +To preserve a meaningful **`last_content_updated`**, **all metadata-only / normalization commits must include the marker**: + +- **Marker**: `META_COMMIT_MARKER` (see `tools/docs_and_notebooks_check.py`) +- **Suggested commit message**: `SUGGESTED_META_COMMIT_MESSAGE` + +When you run `update --write` or `normalize --write`, the tool will: + +- Require `--ack-meta-commit-marker` (guardrail) +- Print a suggested commit message + +> [!WARNING] +> If the marker changes in the future, previous iterations still HAVE to be acknowledged to avoid false positives. + +--- + +## Commands + +### 1) Report (read-only) + +Generate a report (does not modify files): + +```bash +python tools/docs_and_notebooks_check.py report +``` + +Writes (by default): + +- `tmp/docs_nb_checks/docs_nb_checks.json` +- `tmp/docs_nb_checks/docs_nb_checks.md` + +### 2) Check (read-only; may fail) + +Run policy checks. By default, CI will not fail unless allowlists are configured. + +```bash +python tools/docs_and_notebooks_check.py check +``` + +The allowlists live in `tools/docs_and_notebooks_report_config.yml`. +They are currently empty, but can help enforce stricter policies once populated (start empty; "ratchet" later). + +### 3) Update metadata (write mode; explicit intent) + +> [!WARNING] +> `update --write` modifies tracked files. Intended for maintainers (manual), not CI. + +#### 3a) Set `last_content_updated` from git (excluding meta commits) + +```bash +python tools/docs_and_notebooks_check.py update --write --set-content-date-from-git --ack-meta-commit-marker +``` + +#### 3b) Set verification fields (human-controlled) + +```bash +python tools/docs_and_notebooks_check.py update --write --targets docs/page.md examples/JUPYTER/foo.ipynb --set-last-verified today --set-verified-for 3.0.0rc13 --ack-meta-commit-marker +``` + +> Tip: omit `--targets` to operate on all scanned files. + +### 4) Normalize notebooks (explicit churn) + +> [!IMPORTANT] +> Notebook normalization rewrites the notebook JSON into a canonical form. +> As such, it is provided as a separate command. + +Dry-run (shows which files *would* change): + +```bash +python tools/docs_and_notebooks_check.py normalize --targets docs/notebook.ipynb +``` + +Write: + +```bash +python tools/docs_and_notebooks_check.py normalize --write --targets docs/notebook.ipynb --ack-meta-commit-marker +``` + +--- + +## CI integration + +Recommended CI usage: + +- Run `report` on PRs and upload the outputs as artifacts. +- Run `check` once allowlists are populated (start empty to avoid failures). + +> [!IMPORTANT] +> Use `actions/checkout` with `fetch-depth: 0` (or sufficiently deep) so `git log` sees history; shallow clones can cause missing or fallback timestamps. + +Dependencies required for this tool (install in the CI job): + +```bash +pip install pydantic pyyaml nbformat +``` + +--- + +## Troubleshooting + +- If you see `content_date_fallback_to_git_touched`, it usually means one of: + - The checkout history is too shallow, or + - *All* commits touching the file are metadata commits with the marker. + +- If Pydantic raises `class-not-fully-defined` errors, ensure the tool calls `.model_rebuild()` for its models (this is already done in the tool). diff --git a/uv.lock b/uv.lock index 8de940472c..416d4df0e4 100644 --- a/uv.lock +++ b/uv.lock @@ -1078,6 +1078,7 @@ dependencies = [ { name = "pandas", extra = ["hdf5", "performance"] }, { name = "pillow" }, { name = "pycocotools" }, + { name = "pydantic" }, { name = "pyyaml" }, { name = "ruamel-yaml" }, { name = "scikit-image", version = "0.25.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" }, @@ -1146,8 +1147,8 @@ wandb = [ [package.dev-dependencies] dev = [ { name = "coverage" }, + { name = "nbformat" }, { name = "pre-commit" }, - { name = "pydantic" }, { name = "pytest" }, { name = "pytest-cov" }, ] @@ -1176,6 +1177,7 @@ requires-dist = [ { name = "pandas", extras = ["hdf5", "performance"], specifier = ">=2.2,<3" }, { name = "pillow", specifier = ">=7.1" }, { name = "pycocotools" }, + { name = "pydantic", specifier = ">=2,<3" }, { name = "pyside6", marker = "platform_machine == 'x86_64' and sys_platform == 'linux' and extra == 'gui'", specifier = "<6.10" }, { name = "pyside6", marker = "(platform_machine != 'x86_64' and extra == 'gui') or (sys_platform != 'linux' and extra == 'gui')" }, { name = "pyyaml" }, @@ -1213,8 +1215,8 @@ provides-extras = ["gui", "openvino", "docs", "fmpose3d", "tf", "apple-mchips", [package.metadata.requires-dev] dev = [ { name = "coverage" }, + { name = "nbformat", specifier = ">5" }, { name = "pre-commit" }, - { name = "pydantic", specifier = ">=2,<3" }, { name = "pytest" }, { name = "pytest-cov" }, ]