# Episode 0 of shuaishuaicdp/so101-bimanual-v3-all-trajectories

- format: LeRobot v3.0
- robot: so100_so101_bimanual_canonical
- robot family, as the index files it: SO-100 or SO-101, two arms
- kind, as the index files it: recording; teleop (inferred from the name)
- frames per second: 30
- frames in this episode: 149
- task: Fold a T-shirt
- episodes in the dataset: 7,703
- licence: apache-2.0
- dataset page: https://huggingface.co/datasets/shuaishuaicdp/so101-bimanual-v3-all-trajectories

LeRobot v3 keeps many episodes in one file. Download the data file and keep the rows whose episode_index is 0. Each video file holds several episodes back to back: cut it to episode_seconds. Or load it with LeRobot: LeRobotDataset('shuaishuaicdp/so101-bimanual-v3-all-trajectories', episodes=[0]).

Columns of the data file:

- `action`: float32, shape 12 (left_shoulder_pan, left_shoulder_lift, left_elbow_flex, left_wrist_flex, left_wrist_roll, left_gripper, right_shoulder_pan, right_shoulder_lift, right_elbow_flex, right_wrist_flex, right_wrist_roll, right_gripper)
- `observation.state`: float32, shape 12 (left_shoulder_pan, left_shoulder_lift, left_elbow_flex, left_wrist_flex, left_wrist_roll, left_gripper, right_shoulder_pan, right_shoulder_lift, right_elbow_flex, right_wrist_flex, right_wrist_roll, right_gripper)
- `action_normalization_group_index`: int64, shape 1 (group_index)
- `state_normalization_group_index`: int64, shape 1 (group_index)
- `source_repo_index`: int64, shape 1 (repo_index)
- `source_episode_index`: int64, shape 1 (episode_index)
- `source_frame_index`: int64, shape 1 (frame_index)
- `source_task_index`: int64, shape 1 (task_index)
- `source_timestamp`: float32, shape 1 (seconds)
- `source_fps`: float32, shape 1 (frames_per_second)
- `source_camera_frame_mask`: bool, shape 3 (base, left, right)
- `source_frame_is_exact`: bool, shape 1 (is_exact)
- `timestamp`: float32, shape 1

| file | camera | size | this episode | link |
|---|---|---|---|---|
| data |  | 17.8 MB | rows where episode_index == 0 (global frames 0 to 148, 149 frames) | https://huggingface.co/datasets/shuaishuaicdp/so101-bimanual-v3-all-trajectories/resolve/main/data/chunk-000/file-000.parquet |
| video | observation.images.base | 1.4 MB | seconds 0.0 to 4.967 | https://huggingface.co/datasets/shuaishuaicdp/so101-bimanual-v3-all-trajectories/resolve/main/videos/observation.images.base/chunk-000/file-000.mp4 |
| video | observation.images.left | 959.1 KB | seconds 0.0 to 4.967 | https://huggingface.co/datasets/shuaishuaicdp/so101-bimanual-v3-all-trajectories/resolve/main/videos/observation.images.left/chunk-000/file-000.mp4 |
| video | observation.images.right | 771.9 KB | seconds 0.0 to 4.967 | https://huggingface.co/datasets/shuaishuaicdp/so101-bimanual-v3-all-trajectories/resolve/main/videos/observation.images.right/chunk-000/file-000.mp4 |

If your network cannot reach Hugging Face, this site fetches the episode's rows for you, and only that episode's: https://datasets.gurasees.com/episode-data/shuaishuaicdp/so101-bimanual-v3-all-trajectories?episode=0 (parquet) or https://datasets.gurasees.com/episode-data/shuaishuaicdp/so101-bimanual-v3-all-trajectories?episode=0&format=json (JSON). Videos are not relayed.

- next episode: https://datasets.gurasees.com/episodes/shuaishuaicdp/so101-bimanual-v3-all-trajectories.md?episode=1
- last episode: https://datasets.gurasees.com/episodes/shuaishuaicdp/so101-bimanual-v3-all-trajectories.md?episode=7702