--- license: cc-by-4.0 task_categories: - robotics tags: - LeRobot - umi - manipulation - bimanual configs: - config_name: default data_files: data/*/*.parquet --- # Wash up after cooking Bimanual handheld-gripper (UMI-style) demonstrations recorded by Starpilot in a kitchen counter (sink-free) data cell. A human operator holds two instrumented parallel-jaw grippers; every episode is one sub-task of the "Wash up after cooking" cycle. Each episode was checked by a reviewer and its task label confirmed against the video. | Episode | Task | Duration | |---|---|---| | 0 | Wash the pot and mugs | 77.1 s | | 1 | Wipe the counter and hang the towel | 51.7 s | Total: 2 episode(s), 129 s at 30 fps. ## Sensors - 2 handheld parallel-jaw grippers, each with a wrist RGB camera + an aligned depth stream, gripper encoder and IMU - 2 fixed scene cameras - 6-DoF pose of each gripper (`arm*_pose_*` relative, `arm*_pose_abs_*` absolute), gripper opening and width Video streams: `left_wrist_0_camera_rgb_image`, `left_wrist_0_camera_rgb_image_depth`, `right_wrist_0_camera_rgb_image`, `right_wrist_0_camera_rgb_image_depth`, `base_0_camera_rgb_image`, `base_1_camera_rgb_image` (640x480, H.264). `observation.state` (46-dim) and `action` (14-dim: per gripper xyz + rotation vector + grip width) names are in `meta/info.json`. ## Format [LeRobot](https://github.com/huggingface/lerobot) dataset, codebase v3.0 (`robot_type: starpilot_yam_gripper`). Explore clips from this and our other datasets at https://huggingface.co/spaces/starpilot-ai/starpilot-data-sample. More data: https://starpilot.ai