Project Structure

development structure codebase

The project follows a standard Python project layout with separated source, documentation, and configuration directories.

Top-Level Directory

.
├── docker/                 # Docker related files (if any extra config)
├── docs/                   # Documentation (Obsidian Vault)
├── logs/                   # Application logs
├── src/                    # Source code
├── static/                 # Frontend assets (HTML/JS/CSS)
├── checkpoints/            # Model checkpoints
├── data/                   # Dataset directory
├── Dockerfile              # Main Docker build file
├── docker-compose.yml      # Service orchestration
├── Makefile                # Command shortcuts
├── pyproject.toml          # Python dependencies (uv/pip)
└── .env                    # Environment variables

Source Code (src/)

The src/ directory is the python package root.

• api/: FastAPI application and routing.
  • main.py: Entry point.
  • websocket.py: Connection handling.
• core/: Shared utilities and constants.
  • mediapipe_utils.py: Landmark extraction logic.
  • draw_kps.py: Visualization.
• data/: Data loading and processing.
  • mmap_dataset.py: Memory-mapped dataset class.
• modelling/: Neural network components.
  • model.py: Architecture definition.
  • train.py: Training loop.
  • export.py: ONNX export script.

Frontend (static/)

Contains the vanilla JS frontend.

• index.html: Main entry UI.
• live-signs.js: Logic.
• styles.css: Styling.

Related Documentation

• Contributing Guide