README.md

HandPose

Real-time hand tracking and retargeting to ORCA robot hand using MediaPipe and MuJoCo.

If running on macOS, you need to export PYOPENGL_PLATFORM=darwin

Installation

make install 

Or manually:

pip install -e ".[dev]"

To install the HaMeR dependencies, run:

make install-deps

Running the Demos

IK Solution

Uses inverse kinematics (mink) to solve for joint angles:

make ik

Controls:

• q - Quit

Manual Retargeting Solution

Uses direct geometric mapping from hand landmarks to joint angles.

mjpython examples/live_demo.py --camera 0

Controls:

• q - Quit

General structure:

handpose/
├── examples/
│   ├── live_demo_ik.py
│   ├── live_demo_dual.py
│   └── ...
├── handpose/ <--- Library code
│   ├── __init__.py
│   ├── ik_retargeting.py
│   ├── retargeting.py
│   └── tracker.py

Developing

Please keep the codebase well-formatted and linted.

make format
make static-checks