Lyria is a desktop application for real-time webcam tracking. It detects the head, hands, and body using a webcam feed and displays the result as colored bounding boxes and landmarks over the video.
This is an open source project, licensed under the MIT License.
- Real-time head, hand, and body tracking using MediaPipe
- Per-category visibility toggles (head, hands, body)
- Automatic detection of connected cameras, with a selector to switch between them
- Manual and gesture-triggered pose capture, with a countdown before the capture happens
- Short pose sequence recording, with playback in a built-in pose library
- Live log panel showing application and detection activity
- Basic robustness tuning for low-quality webcams (image enhancement, adjusted detection thresholds, landmark smoothing)
The main application window:
Examples of captured poses, showing only the tracking output (head, hands, body):
A short demo video is available at exp_pict/record_exp.mp4.
- Python 3.10 or later
- A webcam
- Windows (camera capture currently relies on a Windows-specific backend)
python -m venv .venv
./.venv/Scripts/python.exe -m pip install -r requirements.txt
The required MediaPipe models are downloaded automatically on first run and cached locally.
./.venv/Scripts/python.exe main.py
Press "Start tracking" to begin. Captured poses are saved locally in poses/; this folder is not tracked by git.
This project is licensed under the MIT License. See LICENSE for details.