PoseAnnotator

PoseAnnotator is a local Python/OpenCV app for annotating person bounding boxes and 2D human poses in images. It is designed for COCO-style workflows and focuses on fast manual annotation with pose-aware visualization.

The repository serves two audiences:

• annotators who need to run the app and produce consistent labels
• developers who need to understand, extend, or maintain the codebase

Start here if you want to use the app for labeling.

• Annotator docs index
• Getting started
• Workflow
• Keyboard shortcuts
• Pose and visibility rules

Minimal commands:

python annotate_bboxes.py <dataset-folder-or-sequence>
python annotate_pose.py <annotation-file-or-sequence>

Start here if you want to install the project locally, understand the structure, or contribute changes.

• Developer docs index
• Project overview
• Setup
• Codebase guide
• Contributing
• Configuration reference

Minimal setup:

pip install -r requirements.txt
pre-commit install

• The app is configured through annotator_config.ini.
• You can pass full paths or shorthand sequence names resolved under sequence_root.
• Pose annotations are written to a suffixed file such as *_manual.json by default.
• Google Drive upload is supported as an optional feature.

Please read docs/dev/contributing.md before making code changes. The short version is: keep the current repo structure, use pre-commit, and update docs when behavior changes.

Maintainers:

• Miroslav Purkrabek - main author and maintainer

Contributors:

• Adela Subrtova
• Fizza Rubab

If you use the tool for research, please consider citing:

@misc{PoseAnnotator2024,
    title={PoseAnnotator},
    author={PoseAnnotator contributors},
    howpublished = {\url{https://github.com/MiraPurkrabek/PoseAnnotator}},
    year={2024}
}