This dataset was created using LeRobot.
Dataset Structure
{
"codebase_version": "v3.0",
"fps": 20,
"features": {
"observation.state": {
"dtype": "float32",
"shape": [
8
],
"names": {
"motors": [
"x",
"y",
"z",
"axis_angle1",
"axis_angle2",
"axis_angle3",
"gripper",
"gripper"
]
}
},
"observation.states.ee_state": {
"dtype": "float32",
"shape": [
6
],
"names": {
"motors": [
"x",
"y",
"z",
"axis_angle1",
"axis_angle2",
"axis_angle3"
]
}
},
"observation.states.joint_state": {
"dtype": "float32",
"shape": [
7
],
"names": {
"motors": [
"joint_0",
"joint_1",
"joint_2",
"joint_3",
"joint_4",
"joint_5",
"joint_6"
]
}
},
"observation.states.gripper_state": {
"dtype": "float32",
"shape": [
2
],
"names": {
"motors": [
"gripper",
"gripper"
]
}
},
"observation.images.image": {
"dtype": "video",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.height": 256,
"video.width": 256,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 20,
"video.channels": 3,
"has_audio": false,
"video.g": 2,
"video.crf": 30,
"video.preset": 12,
"video.fast_decode": 0,
"video.video_backend": "pyav",
"video.extra_options": {}
}
},
"observation.images.wrist_image": {
"dtype": "video",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.height": 256,
"video.width": 256,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 20,
"video.channels": 3,
"has_audio": false,
"video.g": 2,
"video.crf": 30,
"video.preset": 12,
"video.fast_decode": 0,
"video.video_backend": "pyav",
"video.extra_options": {}
}
},
"observation.images.frontview_image": {
"dtype": "video",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.height": 256,
"video.width": 256,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 20,
"video.channels": 3,
"has_audio": false,
"video.g": 2,
"video.crf": 30,
"video.preset": 12,
"video.fast_decode": 0,
"video.video_backend": "pyav",
"video.extra_options": {}
}
},
"observation.images.sideview_image": {
"dtype": "video",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.height": 256,
"video.width": 256,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 20,
"video.channels": 3,
"has_audio": false,
"video.g": 2,
"video.crf": 30,
"video.preset": 12,
"video.fast_decode": 0,
"video.video_backend": "pyav",
"video.extra_options": {}
}
},
"action": {
"dtype": "float32",
"shape": [
7
],
"names": {
"motors": [
"x",
"y",
"z",
"axis_angle1",
"axis_angle2",
"axis_angle3",
"gripper"
]
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
},
"total_episodes": 253,
"total_frames": 55988,
"total_tasks": 10,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 1e-06,
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"robot_type": "kinova3",
"splits": {
"train": "0:253"
}
}
Citation
BibTeX:
[More Information Needed]
TsFile conversion
This repository is a compact Apache TsFile conversion of the source LeRobot v3.0 Kinova3 dataset. The converted train split has 55,988 rows, 253 episodes/devices, 10 tasks, one source Parquet shard, and 20 fps sampling.
Attribution and videos
The source dataset is published under LSY-lab / Learning Systems Lab and was uploaded by Maximilian Christof (Hugging Face user max-chr).
- Source: https://huggingface.co/datasets/LSY-lab/libero_plus_goal_texture_all_kinova3_failures
- License: Apache-2.0
- Source revision: 7693e9ab1ca5650b4538658511a8cc555ae6091b
- The source card does not provide a dataset-specific paper citation.
- Source video layout: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
- Video streams: observation.images.image, observation.images.wrist_image, observation.images.frontview_image, observation.images.sideview_image.
- Online inventory: 253 MP4s per stream, 1,012 videos total, about 765,453,271 bytes (730 MiB).
- Videos are not included here; use https://huggingface.co/datasets/LSY-lab/libero_plus_goal_texture_all_kinova3_failures/tree/main/videos
Converted schema
| Group | Columns | Role/type |
|---|---|---|
| Time | Time | TIME INT64 milliseconds |
| TAG/device | episode_index, task_index | original TsFile TAG columns |
| Scalar FIELD | frame_index, sample_index | INT64 FIELD; sample_index is source index |
| FLOAT FIELD | observation_state_0..7 | source observation.state[8] |
| FLOAT FIELD | observation_states_ee_state_0..5 | source observation.states.ee_state[6] |
| FLOAT FIELD | observation_states_joint_state_0..6 | source observation.states.joint_state[7] |
| FLOAT FIELD | observation_states_gripper_state_0..1 | source observation.states.gripper_state[2] |
| FLOAT FIELD | action_0..6 | source action[7] |
Time = round(timestamp * 1000) ms and restarts per episode. Source timestamp is dropped as redundant; index is renamed to sample_index; frame_index is retained for video alignment. Vector fields are flattened row-major with periods replaced by underscores. No rows or numeric state/action fields are dropped.
FLOAT/DOUBLE use GORILLA + LZ4; INT32/INT64 and Time use TS_2DIFF + LZ4; BOOLEAN uses RLE + LZ4; TAGs use TsFile table/device storage. Camera payload features are omitted because their MP4 files remain in the original dataset.
Converted file: data/lsy_lab_libero_plus_goal_texture_all_kinova3_failures.tsfile
from tsfile import TsFileReader
reader = TsFileReader("data/lsy_lab_libero_plus_goal_texture_all_kinova3_failures.tsfile")
print(reader.get_all_table_schemas())
Local inspection and validation reports are excluded from the upload set.
- Downloads last month
- 66