RoboMoCap dataset released: 200+ hours of 120Hz whole-body mocap for humanoid robot learning
DJiafei · x · 2026-11-09
RoboMoCap, a high-fidelity whole-body motion dataset for humanoid learning, is now public on Hugging Face with an accompanying paper.
- Scale: 200+ hours from 14 participants
- Capture: 6 head-mounted + 2 wrist cameras, 120 Hz optical whole-body mocap, IMU + magnetometer, hand poses, tracked 3D objects — multimodal observations aligned with whole-body motion
- 290+ actions across 8 families: locomotion, loco-manipulation, posture & dynamics, stationary manipulation, dynamic recovery, human–human interaction, environment-coupled motion, long-horizon tasks
- Positioning: most egocentric datasets focus on hand–object interactions; RoboMoCap puts locomotion and loco-manipulation at the center
- Comparison: ACE-Data-0 150h (household), Nymeria 300h (inertial mocap), EgoHTR 1.37h (terrain traversal)
More from Embodied
- DefFlow: Robots Manipulate Ropes, Cloth into Precise 3D Goal Shapes via Point Flows — drsrinathsridha · 2026-11-09
- Lamp's onboarding guide gets turned into a comic book — dee_hw · 2026-11-09
- No vision, no tactile sensors: humanoid dexterity from proprioception alone at CoRL — Jan_R_Peters · 2026-11-09
- Geometric abstraction cuts robot agent VLM queries 52.8% and speeds manipulation 33% — siddkaramcheti · 2026-11-09
- Autonomous's $299 Intern 2 is a plug-and-play desktop AI agent in iMessage — dee_hw · 2026-11-09
- NVIDIA unveils GR00T 2 world action model at CoRL: 71% success with one demo — NVIDIA Blog · 2026-11-09