ReCAP

Motion capture retargeting library for humanoid robots

Features

• Easy to use inerface for solving IK for humanoid robots
• MJCF model editor for adding new reference frames to the robot model
• CMU and AMASS mocap datasets support. Library can be extended to support other datasets
• Optimization based SMPLH parameter estimation
• Torch based forward kinematics for MJCF models

References

• robot_descriptions - Extensive collection of MJCF robot models
• pink - Awesome library for diffrential IK
• quaternion - Convenient library for quaternion operations
• SMPL - Vertex based linear human body model
• AMASS - Aggregated motion capture dataset utilizing SMPL(XH) models
• CMU - CMU motion capture dataset

Installation

Dependencies are listed in pyproject.toml. However numpy-quaternion needs to be forced to be compiled. For more info refer to the numpy-quaternion README.

python -m pip install --upgrade --force-reinstall numpy-quaternion

Before installing the package you might want to create a new conda environment:

conda create -n recap python=3.10

Then activate the environment:

conda activate recap

Clone the repository:

git clone https://github.com/RumblingTurtle/recap.git
cd recap

Install the package:

pip install .

Or if you are planning to extend the library, you can install the package locally using pip:

pip install -e .

Contributing

Feel free to contribute to the library by opening a pull request or an issue.

Documentation

• Downloading datasets
• Adding new datasets
• Adding new robot models
• Solving IK
• Examples