RoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation
Paper • 1811.02790 • Published
codebase_version string | robot_type string | total_episodes int64 | total_frames int64 | total_tasks int64 | total_videos int64 | total_chunks int64 | chunks_size int64 | fps int64 | splits dict | data_path string | features dict | video_path_original string | tsfile_conversion dict |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
v2.1 | sawyer | 1,796 | 168,423 | 3 | 1,796 | 2 | 1,000 | 10 | {
"train": "0:1796"
} | data/roboturk.tsfile | {
"Time": {
"dtype": "int64",
"shape": [
1
],
"tsfile_role": "TIME",
"unit": "ms"
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"tsfile_role": "TAG"
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"tsfile_role": "TAG"
},
... | videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4 | {
"source_dataset": "IPEC-COMMUNITY/roboturk_lerobot",
"source_data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"converted_data_path": "data/roboturk.tsfile",
"table_name": "roboturk",
"granularity": "merged",
"time_precision": "ms",
"time_mapping": {
"source": "auto",... |
RoboTurk (Sawyer) (TsFile)
Apache TsFile version of IPEC-COMMUNITY/roboturk_lerobot.
Overview
RoboTurk is a large-scale robot manipulation dataset of human teleoperation demonstrations collected on a Sawyer robot arm, distributed here as part of the Open X-Embodiment collection in the LeRobot v2.1 format. Each record is one control frame within a demonstration episode, holding the robot's 8-D proprioceptive state and the 7-D end-effector action.
- Scale: 1,796 episodes, 168,423 frames total (≈ 94 frames per episode on average).
- Sampling rate: 10 fps.
- Robot: Sawyer arm.
- Tasks: 3 —
object search,layout laundry,create tower(seetask_index).
Schema (TsFile structure)
- Time (INT64, milliseconds) —
round(timestamp * 1000), restarting at 0 for each episode. - episode_index (TAG) — source episode id (0–1795).
- task_index (TAG) — source task id (0 = object search, 1 = layout laundry, 2 = create tower).
- frame_index (FIELD, INT64) — frame position within the episode.
- sample_index (FIELD, INT64) — the source global
indexcolumn, renamed. - observation_state_0 .. observation_state_7 (FIELD, FLOAT) — the 8-D state (motor_0..7), flattened.
- action_0 .. action_6 (FIELD, FLOAT) — the 7-D action (x, y, z, roll, pitch, yaw, gripper), flattened.
All 1,796 episodes share a single .tsfile; episode_index and task_index are
the device (TAG) dimensions. Query one episode with WHERE episode_index=0, or a
whole task with WHERE task_index=1.
Vector columns are flattened keeping the source name (observation.state →
observation_state_0..7, action → action_0..6). The source timestamp
column is dropped because it equals Time / 1000 seconds; frame_index is kept.
Usage
Read the .tsfile files with the Apache TsFile Java or Python SDK.
Source & license
- Original dataset: https://huggingface.co/datasets/IPEC-COMMUNITY/roboturk_lerobot
- Author / publisher: IPEC-COMMUNITY (Hugging Face); original RoboTurk dataset.
- Paper: RoboTurk — A Crowdsourcing Platform for Robotic Skill Learning through Imitation (https://arxiv.org/abs/1811.02790)
- Part of the Open X-Embodiment collection.
- Camera video stream (
observation.images.front_rgb, 448×448 RGB @ 10 fps) is not included in this repository; see the original dataset'svideos/directory for it. - License: apache-2.0.
- Downloads last month
- 66