1 episodes · 30 fps · 4 cameras · 640×400 av1
This dataset was created using LeRobot.
Dataset Description
A 30 s egocentric bimanual manipulation clip from a head-mounted stereo rig with two hand-held UMI-style grippers
(each with a wrist camera + IMU and an ArUco marker cube), recorded with Starpilot's NanoPi VIO rig.
It covers 80-110 s of recording 20261006_080130 ("test2"): 900 frames at 30 fps.
Poses come from our own pipeline: stereo visual odometry for the head, stereo-triangulated ArUco cubes for the grippers (gaps bridged with each gripper's wrist IMU, orientation fused gyro + markers), and finger-marker tracking in the wrist cameras for gripper opening. This is a test release of the format; expect rough edges.
Head-mounted VIO rig recording converted to the LeRobotDataset v3.0 format (codebase_version: "v3.0"). It was written with lerobot==0.6.1 (LeRobotDataset.create / add_frame / save_episode / finalize) and loads with LeRobotDataset("starpilot-ai/mobile_umi_test").
- Source: recording
20261006_080130(label "test2"), cropped to 80-110 s. 1 episode, 900 frames, 30 fps, 30.0 s. - Converter:
vio_rig/to_lerobot.py(internal). Rig-specific notes are inmeta/vio_rig.json. They are not ininfo.json, because lerobot drops unknown keys there. - The task string is the recording's label (e.g. "test2").
Layout (standard v3.0)
meta/info.json meta/stats.json meta/tasks.parquet meta/episodes/chunk-000/file-000.parquet
data/chunk-000/file-000.parquet
videos/observation.images.<cam>/chunk-000/file-000.mp4 (AV1/libsvtav1, yuv420p, crf 30, GOP 2)
Frames and timing
The stereo camera is the master timeline, with one dataset frame per stereo frame (synced_frames.csv). The monos are hardware-triggered by the stereo strobe. timestamp = frame_index / 30. exposure_t is the measured stereo exposure time (host clock) relative to the first frame.
Coordinate frames and units
- world: gravity-aligned, per episode, in metres: z up (gravity measured by the two wrist IMUs, which agree within ~0.7 deg), x = the head camera's viewing direction at frame 0 flattened onto the horizontal, origin = the head (stereo_L) camera's position at frame 0. Each episode has its own world.
- Smoothing: positions of the head and grippers get the lightest Gaussian smoothing (sigma 1 frame, +-2 frames) within continuous stretches. Gripper orientation is fused: the wrist gyro carries the fast motion, the ArUco cube the slow absolute part (outlier faces rejected), so there are no frame-to-frame orientation jumps.
- head: the current stereo_L camera frame (from
vo_poses.csv, stereo VO). - Poses are
x y z qx qy qz qw(scipy/ROS quaternion order, normalised to qw >= 0). - Gripper poses are the pose of the marker cube centre (
grippers_fused.csv).
Features
| key | shape | meaning |
|---|---|---|
| observation.images.stereo_left / stereo_right | 400x640x3 video | head stereo pair with the 160 px code strip removed (1920x1200 to 640x400) |
| observation.images.wrist_A / wrist_B | 400x640x3 video | wrist cameras: mono_1 on gripper A, mono_2 on gripper B |
| observation.state | 23 | head pose in world (7), A pose in head frame (7), A opening_norm (1), B pose in head frame (7), B opening_norm (1) |
| observation.head.pose_world | 7 | head (stereo_L) pose in world |
| observation.gripper_{A,B}.pose_world | 7 | gripper pose in world |
| observation.gripper_{A,B}.pose_head | 7 | gripper pose in the current head frame |
| observation.gripper_{A,B}.opening | 3 | opening_norm (0..1), opening_mm, opening_deg (from gripper_opening.csv) |
| observation.imu_{A,B} | 6 | mean gx gy gz (rad/s) and ax ay az (m/s^2) of the wrist-camera IMU over the frame interval. Raw IMU axes, no bias removal, gyro scale 28.65 LSB/(deg/s) |
| action | 20 | per gripper (A then B), absolute: the gripper pose at t+1 in the world frame (position, 3, m; rotation as 6D = first two rows of the matrix, 6) and its opening_norm at t+1 (1) |
| valid | 13 | 1 = measured, 0 = filled. Order: head_pose, gripper_A_pose, gripper_A_opening, gripper_B_pose, gripper_B_opening, img_stereo_left, img_stereo_right, img_wrist_A, img_wrist_B, imu_A, imu_B, action_A, action_B |
| pose_source | 2 | per gripper: 0 = missing, 1 = ArUco markers, 2 = IMU fill (from grippers_fused source) |
| head_vo_inliers | 1 | VO inlier count (a quality indicator) |
| exposure_t | 1 | seconds since the first exposure |
| timestamp, frame_index, episode_index, index, task_index | standard LeRobot columns |
LeRobot policies use the observation.* keys as inputs and action as the output. valid, pose_source, head_vo_inliers and exposure_t are metadata. Their names were chosen so that policies ignore them.
Missing-data convention
Every feature is dense. If a value is unknown for a frame, the converter does the following:
- It holds the last valid value (forward fill).
- It back-fills a leading gap from the first valid value.
- If a quantity is never valid in the episode, it writes zeros, or the identity pose for poses.
- It sets the matching
validentry to 0.
An action is valid when the gripper pose is valid at t+1. The last frame's action is always invalid. Mask losses with valid. Openings that the opening pipeline marked ...;interp or fixed_from_median count as valid.
Coverage in this clip (valid / 900): head 900, gripper_A pose 900, gripper_A opening 892, gripper_B pose 900, gripper_B opening 860, both IMUs 900, all four cameras 900.
- Homepage: [More Information Needed]
- Paper: [More Information Needed]
- License: apache-2.0
Dataset Structure
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"observation.images.stereo_left": {
"dtype": "video",
"shape": [
400,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 400,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.fps": 30,
"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": {},
"is_depth_map": false
}
},
"observation.images.stereo_right": {
"dtype": "video",
"shape": [
400,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 400,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.fps": 30,
"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": {},
"is_depth_map": false
}
},
"observation.images.wrist_A": {
"dtype": "video",
"shape": [
400,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 400,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.fps": 30,
"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": {},
"is_depth_map": false
}
},
"observation.images.wrist_B": {
"dtype": "video",
"shape": [
400,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 400,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.fps": 30,
"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": {},
"is_depth_map": false
}
},
"observation.state": {
"dtype": "float32",
"shape": [
23
],
"names": [
"head_x",
"head_y",
"head_z",
"head_qx",
"head_qy",
"head_qz",
"head_qw",
"A_x",
"A_y",
"A_z",
"A_qx",
"A_qy",
"A_qz",
"A_qw",
"A_opening",
"B_x",
"B_y",
"B_z",
"B_qx",
"B_qy",
"B_qz",
"B_qw",
"B_opening"
]
},
"observation.head.pose_world": {
"dtype": "float32",
"shape": [
7
],
"names": [
"x",
"y",
"z",
"qx",
"qy",
"qz",
"qw"
]
},
"observation.gripper_A.pose_world": {
"dtype": "float32",
"shape": [
7
],
"names": [
"x",
"y",
"z",
"qx",
"qy",
"qz",
"qw"
]
},
"observation.gripper_A.pose_head": {
"dtype": "float32",
"shape": [
7
],
"names": [
"x",
"y",
"z",
"qx",
"qy",
"qz",
"qw"
]
},
"observation.gripper_A.opening": {
"dtype": "float32",
"shape": [
3
],
"names": [
"opening_norm",
"opening_mm",
"opening_deg"
]
},
"observation.imu_A": {
"dtype": "float32",
"shape": [
6
],
"names": [
"gx",
"gy",
"gz",
"ax",
"ay",
"az"
]
},
"observation.gripper_B.pose_world": {
"dtype": "float32",
"shape": [
7
],
"names": [
"x",
"y",
"z",
"qx",
"qy",
"qz",
"qw"
]
},
"observation.gripper_B.pose_head": {
"dtype": "float32",
"shape": [
7
],
"names": [
"x",
"y",
"z",
"qx",
"qy",
"qz",
"qw"
]
},
"observation.gripper_B.opening": {
"dtype": "float32",
"shape": [
3
],
"names": [
"opening_norm",
"opening_mm",
"opening_deg"
]
},
"observation.imu_B": {
"dtype": "float32",
"shape": [
6
],
"names": [
"gx",
"gy",
"gz",
"ax",
"ay",
"az"
]
},
"action": {
"dtype": "float32",
"shape": [
20
],
"names": [
"A_x",
"A_y",
"A_z",
"A_rot6d_r00",
"A_rot6d_r01",
"A_rot6d_r02",
"A_rot6d_r10",
"A_rot6d_r11",
"A_rot6d_r12",
"A_opening_next",
"B_x",
"B_y",
"B_z",
"B_rot6d_r00",
"B_rot6d_r01",
"B_rot6d_r02",
"B_rot6d_r10",
"B_rot6d_r11",
"B_rot6d_r12",
"B_opening_next"
]
},
"valid": {
"dtype": "float32",
"shape": [
13
],
"names": [
"head_pose",
"gripper_A_pose",
"gripper_A_opening",
"gripper_B_pose",
"gripper_B_opening",
"img_stereo_left",
"img_stereo_right",
"img_wrist_A",
"img_wrist_B",
"imu_A",
"imu_B",
"action_A",
"action_B"
]
},
"pose_source": {
"dtype": "float32",
"shape": [
2
],
"names": [
"gripper_A",
"gripper_B"
]
},
"head_vo_inliers": {
"dtype": "float32",
"shape": [
1
],
"names": [
"inliers"
]
},
"exposure_t": {
"dtype": "float32",
"shape": [
1
],
"names": [
"seconds_since_episode_start"
]
},
"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": 1,
"total_frames": 900,
"total_tasks": 1,
"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": "vio_rig_handheld_bimanual",
"splits": {
"train": "0:1"
}
}
Citation
BibTeX:
[More Information Needed]
- Downloads last month
- -