as54/lerobot_ego_grasp_test1

robox/ego-grasp-1783507905

Egocentric hand-manipulation dataset in LeRobot v3.0 format, produced by the Robox export pipeline from real on-device hand-tracking recordings. It is structured for direct use with LeRobot (ACT / Diffusion / pi0-style policies).

Dataset Summary

• Episodes: 1276
• Frames: 186255
• Tasks: 1101
• FPS: 30
• Robot type: robox_ego_hand
• Codebase version: v3.0
• State / action dim: 15

State / action components:

wrist_x, wrist_y, wrist_z, wrist_qx, wrist_qy, wrist_qz, wrist_qw, finger_curl_thumb, finger_curl_index, finger_curl_middle, finger_curl_ring, finger_curl_pinky, pinch_thumb_index, pinch_thumb_middle, pinch_thumb_ring

Features

Key	dtype	shape
observation.images.front	video	[480, 640, 3]
observation.state	float32	[15]
action	float32	[14]
timestamp	float32	[1]
frame_index	int64	[1]
episode_index	int64	[1]
index	int64	[1]
task_index	int64	[1]

Structure

meta/info.json          # schema, fps, feature spec (canonical)
meta/tasks.parquet      # task_index -> natural-language task
meta/episodes/*.parquet # per-episode index, length, video timestamps
meta/stats.json         # per-feature statistics
meta/robox_export.json  # licence + provenance (Robox sidecar)
meta/robox_validation.json  # structural validation report
data/chunk-000/*.parquet    # per-frame state / action / timestamp
videos/observation.images.front/chunk-000/*.mp4  # RGB, one shard per chunk

Episodes are packed into shared chunk shards (one .mp4 and one .parquet per chunk); per-episode boundaries are addressed by the row ranges and video timestamps in meta/episodes.

meta/info.json (full canonical schema)

{
    "codebase_version": "v3.0",
    "fps": 30,
    "features": {
        "observation.images.front": {
            "dtype": "video",
            "shape": [
                480,
                640,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "video.height": 480,
                "video.width": 640,
                "video.codec": "h264",
                "video.pix_fmt": "yuv420p",
                "video.fps": 30,
                "video.channels": 3,
                "has_audio": false,
                "video.g": 2,
                "video.crf": 30,
                "video.preset": null,
                "video.fast_decode": 0,
                "video.video_backend": "pyav",
                "video.extra_options": {},
                "is_depth_map": false
            }
        },
        "observation.state": {
            "dtype": "float32",
            "shape": [
                15
            ],
            "names": [
                "wrist_x",
                "wrist_y",
                "wrist_z",
                "wrist_qx",
                "wrist_qy",
                "wrist_qz",
                "wrist_qw",
                "finger_curl_thumb",
                "finger_curl_index",
                "finger_curl_middle",
                "finger_curl_ring",
                "finger_curl_pinky",
                "pinch_thumb_index",
                "pinch_thumb_middle",
                "pinch_thumb_ring"
            ]
        },
        "action": {
            "dtype": "float32",
            "shape": [
                14
            ],
            "names": [
                "d_wrist_x",
                "d_wrist_y",
                "d_wrist_z",
                "d_wrist_rotvec_x",
                "d_wrist_rotvec_y",
                "d_wrist_rotvec_z",
                "d_finger_curl_thumb",
                "d_finger_curl_index",
                "d_finger_curl_middle",
                "d_finger_curl_ring",
                "d_finger_curl_pinky",
                "d_pinch_thumb_index",
                "d_pinch_thumb_middle",
                "d_pinch_thumb_ring"
            ]
        },
        "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": 1276,
    "total_frames": 186255,
    "total_tasks": 1101,
    "chunks_size": 1000,
    "data_files_size_in_mb": 100,
    "video_files_size_in_mb": 200,
    "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": "robox_ego_hand",
    "splits": {
        "train": "0:1276"
    }
}

Licence & Provenance

• Licence: see meta/robox_export.json
• Export id: beb4b060-19da-47f6-a7da-c1aaf68ab655
• Source recordings: 1000

Full licensing, buyer and source-recording provenance are in meta/robox_export.json. This dataset is distributed under its stated commercial terms — not an open-source licence — despite the Hub license: other tag.

Created with

This dataset was created using LeRobot and the Robox LeRobot v3 export worker.