license: cc-by-nc-sa-4.0
tags:
- lerobot
- robotics
- feeding
language:
- en
task3_phase2
Task 3: Spoon feeding the mannequin. Visually selected feeding recordings from two source disks. Unrelated thermal-pad placement and tray-transport recordings are excluded. Source success and failure labels are retained; the task description supplied by the user overrides stale source task metadata.
Task: Pick up the spoon, feed the mannequin, and place the spoon back in the bowl.
LeRobot v2.1, 177 episodes, 124872 frames, 20 Hz, three H.264 cameras at their original resolutions.
Source success/failure labels are retained in annotation.human.validity. State has 42 dimensions and action has 17 dimensions, including the original converter configuration's synthetic spine height of 434.0. Joint velocities are not included. The supplied converter's next.reward convention (1 on the last frame of every episode) is retained; validity carries task success/failure.
The videos are re-encoded at CQ 18 to align with the fixed 20 Hz robot-data time grid. Head camera approximately undistorted using factory parameters from a different ZED Mini, SN17064700, left eye, alpha=0. The recording camera serial and captured eye were not independently verified. All output pixels sample valid recorded pixels. Uniform focal scaling removes invalid borders, with a top-anchored crop to retain the mannequin mouth. No generated pixels or anisotropic stretching. Wrist camera geometry is unchanged. Native image dimensions and square pixels are retained. This is not a verified per-device calibration. The input and output camera parameters and valid-pixel mask are included under meta/. The common timeline includes only times covered by all required streams. Source MCAP files and checksums are retained separately.
from lerobot.datasets.lerobot_dataset import LeRobotDataset
dataset = LeRobotDataset("ebim-benchmark/task3_phase2", root="path/to/local/dataset", video_backend="pyav")