Hyundai Uiwang — Diffusion Policy (FlowMatch), RIGHT arm
LeRobot Diffusion Policy trained on the Hyundai Uiwang right-arm manipulation
dataset (118 episodes / 77,490 frames, 30 Hz) with the FlowMatch scheduler
(num_inference_steps=1). Robot: HDRB + INSPIRE-RH56.
I/O
- in
observation.images.front_rgb(H,W,3 uint8),observation.images.wrist_rgb(H,W,3 uint8),observation.state(26 float) - out
action(26 float): target arm (6) + hand (20) joints @ 30 Hz
Quick start
pip install "lerobot @ git+https://github.com/snu-hyundai-robot-proj/lerobot.git@main"
huggingface-cli download Ngseo/hyundai-uiwang-right-flowmatch inference_example.py --local-dir .
python inference_example.py --model-id Ngseo/hyundai-uiwang-right-flowmatch --device cuda
Serving (HTTP, Docker): https://github.com/snu-hyundai-robot-proj/visuomotor-policy-inference
- Downloads last month
- 3