Poseless is a project that enables direct estimation of robot hand joint angles from human hand input without requiring complex 3D skeleton structure or pose estimation pipelines. This simplified approach provides a more efficient way to map human hand movements to robotic hand control.
The system directly maps human hand input to robot joint angles, bypassing the need for intermediate 3D pose estimation. This approach offers:
- Reduced computational complexity
- Lower latency
- Direct joint angle mapping
- Simplified implementation
- Python 3.x
- Required dependencies please check
requirements.txt
# Clone the repository
git clone https://github.com/yourusername/poseless.git
cd poseless
# Install dependencies (to be added)
sudo apt update
sudo apt install libegl1 libegl-dev
ldconfig -p | grep libEGL #check if
pip install -r requirements.txt
To generate samples, use the following command:
export PYOPENGL_PLATFORM=egl
export MUJOCO_GL=egl # Also set this.
python gen_multithread.py <number_to_generate> --dataset_name="your-username/hand-poses-dataset" --num_test_samples <num_test_samples>