CoolFace
Datasetpublic

robot-lev/rh20t_cfg4

rh20t_cfg4 — RH20T → LeRobot v3 (ur5) Unofficial LeRobot Dataset v3 reformatting of an RH20T config. RGB only (no depth). 2182 episodes / 1272711 frames, fps 10, robot_type=ur5, 9 camera streams. Attribution RH20T: Fang et al., RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot, arXiv:2307.00595. Project: https://rh20t.github.io · API: https://github.com/rh20t/rh20t_api Unofficial community reformatting — all credit to the RH20T… See the full description on the dataset page: https://huggingface.co/datasets/robot-lev/rh20t_cfg4.

sourceHugging Faceotherupdated 3mo agoView on Hugging Face
0likes990downloads
Dataset Card

rh20t_cfg4 — RH20T → LeRobot v3 (ur5)

Unofficial LeRobot Dataset v3 reformatting of an RH20T config. RGB only (no depth). 2182 episodes / 1272711 frames, fps 10, robot_type=ur5, 9 camera streams.

Attribution

RH20T: Fang et al., RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot, arXiv:2307.00595. Project: https://rh20t.github.io · API: https://github.com/rh20t/rh20t_api Unofficial community reformatting — all credit to the RH20T authors. Reformatting code: https://github.com/lvjonok/rh20t-lerobot-port

License — mirrors RH20T's per-scene dual license

RH20T is dual-licensed by scene; this is preserved here:

  • —Episodes whose source folder contains scene_0001–scene_0005 → CC BY-SA 4.0 (commercial use allowed; attribution + ShareAlike).
  • —Episodes whose source folder contains scene_0006–scene_0010 → CC BY-NC 4.0 (non-commercial only — commercial use of these episodes, or of models trained on them, is not allowed).

Each episode's source folder is recorded in meta/rh20t_episodes.json (folder field); use it to determine the applicable license per episode.

Sensitive content (per RH20T)

Contains volunteer-recorded human–robot interactions that may include faces (video) and voices (audio sidecars). Exercise care to avoid inspecting or sharing sensitive content; utilize the dataset solely for model-training purposes.

Schema

observation.state (eepose 7 + joint 6 + gripper 1) · `observation.force`/`torque` (3+3) · `observation.robotft (6) · action (next ee_pose 7 + gripper 1) · meta.rating · 9 × observation.images.cam_<serial> (video). See DECISIONS.md` in the port repo for every assumption.