Access requires agreeing to the Pantheon Commercial Use License
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This dataset is released under the Pantheon Commercial Use License, Version 1.0 (see LICENSE.md). It may be used for any lawful purpose, including commercial use and training models, subject to the license. You may not sell or redistribute the data itself for a fee, and you may not attempt to identify anyone in it.
Questions: data@pantheon.inc
Log in or Sign Up to review the conditions and access this dataset content.
Pantheon DiverseUMI (100h Sample)
This dataset holds the videos, poses and actions of every episode, and their labels are in their own dataset, DiverseUMI-sample100-annotations.
Project page: pantheon.inc/diverse-umi
Pantheon DiverseUMI (100h Sample) contains 99.6 hours of bimanual tabletop manipulation, recorded with two handheld UMI grippers and a fisheye camera above the table (a 360° dual-fisheye camera in some recordings). It has 12,219 episodes from 1,134 recordings. Each episode is one task, such as stacking two blocks or passing a ball between the grippers. Every episode has three synchronized videos (the top camera and both wrist cameras), the pose and actions of both wrists at every frame, and dense labels of what the operator does.
Operators worked freely with everyday objects. Tasks in one recording can overlap in time, so 156 pairs of episodes share some frames. Of the 12,219 tasks, 11,563 succeed, 441 fail and 215 are partially completed. Episodes run from 1s to 1031s, with a median of 10s. The tasks cover 179 skills, led by pick and place (5,826), stack (894), hand over (528) and stand up (487).
Formats
The dataset is published as three datasets, which share episode_index and episode_id:
- DiverseUMI-sample100, in the LeRobot dataset format v3.0, which
LeRobotDatasetloads (this dataset) - DiverseUMI-sample100-lance, the same data in LeRobot's Lance format, which
LeRobotDatasetalso loads and which is laid out for training straight from the Hub - DiverseUMI-sample100-annotations, the labels of every episode, one line per episode
Changelog
Every update to the dataset is listed in the CHANGELOG.
Full dataset and collaborations
For access to more of DiverseUMI through a research or data collaboration, reach out to data@pantheon.inc. We work with both commercial and academic labs.
License
This dataset is released under the Pantheon Commercial Use License, Version 1.0. In short:
- You may use it for any lawful purpose, including commercial products and research.
- You may train, fine-tune and evaluate models on it. You own the models you train and what they generate, and you can deploy, release or sell them on your own terms.
- You may not sell, rent or license the data itself, or anything that reproduces it, or offer it as a dataset or data product.
- You may not use it to build a dataset or data product that competes with this dataset or with Pantheon's data products. Training and selling models is not affected.
- You may not try to identify anyone in the videos, extract biometric data, or combine the data with other data in a way that could identify someone.
If you share the data or a copy derived from it (not a trained model), include the license and this attribution:
Pantheon DiverseUMI (100h Sample), by Pantheon Industries Inc., licensed under the Pantheon Commercial Use License, Version 1.0. https://huggingface.co/datasets/Pantheon-inc/DiverseUMI-sample100
Technical details
Expand each section for the files, the cameras, every field, and how the actions were made.
Files
meta/info.json the features, the frame rate and where every file is
meta/stats.json statistics of every feature
meta/tasks.parquet every task, by task_index
meta/episodes/ one row per episode: its length, tasks and where its frames and videos are
meta/episode_info.jsonl one line per episode with its metadata (below, under Episodes)
data/ one row per frame, with every feature (below, under Features)
videos/observation.images.exo/ top camera (called exo in the labels)
videos/observation.images.left_wrist/ left wrist camera
videos/observation.images.right_wrist/ right wrist camera
The three videos run at 30 fps and start on the same frame, and t, the frame's time from the start of the episode, equals LeRobot's timestamp. Frame k of an episode is its row with frame_index k, and the per-step features in that row are the step from
frame k to k+1. Episodes are numbered by episode_index in the order of their episode_id. Each episode is
named <recording_id>__t<NN>, where NN is the task's index in its recording (two or more digits).
Cameras
The wrist videos are 640 x 480, and the top camera's video is 784 x 504 in every episode. All three are AV1, with a keyframe every 8 frames. The top camera records a fisheye image, and its video is a flat (rectilinear) view of the table taken from it, rendered at a size that depends on the camera and how it was mounted and then resized to 784 x 504:
| Rendered size | Episodes | Source |
|---|---|---|
| 796 x 510 | 4,112 | lowered camera, fisheye source |
| 512 x 328 | 3,520 | fisheye source |
| 1362 x 876 | 2,250 | fisheye source |
| 908 x 584 | 2,161 | dual fisheye source |
| 696 x 448 | 176 | dual fisheye source |
From 2026-09-28 on the top camera was mounted lower, at 55cm above the table instead of 100cm. In those 4,112 episodes exo_camera_lowered is true, and the
view looks 20° down instead of 30° and keeps a wider part of the table in view.
exo_view in meta/episode_info.jsonl gives the exact view of every episode, so it can be reproduced from the fisheye recording:
| Key | Meaning |
|---|---|
source_lens, source_fov_deg |
the top camera's lens (fisheye, or dual fisheye for a 360° camera) and the field of view of one lens |
projection, fov_deg |
rectilinear, with this horizontal and vertical field of view of the full pane in degrees, each set on its own, so a pixel need not span the same angle across and down |
pitch_deg |
the view's pitch in degrees; negative looks down |
pane |
width and height of the full rectilinear view in pixels |
crop |
x, y, width and height of the part of the pane that is kept |
size |
width and height the view was rendered at, before it was resized to 784 x 504 |
Episodes
meta/episode_info.jsonl has one line per episode, in episode_index order.
| Field | Meaning |
|---|---|
episode_index |
the episode's number in this dataset and in its companions |
episode_id |
<recording_id>__t<NN> |
recording_id |
the recording the episode was cut from |
task_index |
the task's position among the tasks found in the recording, counting from 0 |
task_count |
how many tasks were found in the recording. Not every task has an episode, so some task indexes are missing |
task |
what the operator does in this episode |
skill |
the task's main skill |
outcome |
success, failure or partial |
start_in_recording_s, end_in_recording_s |
where the episode starts and ends in its recording, in seconds |
seconds |
the episode's length |
min_wrist_translation_valid_share |
of the two wrists, the lower share of steps with a valid translation. It is at least 0.5 in every episode |
gripper_version |
v3, or v2 for the 2 recordings (27 episodes) made with the older gripper, which has black slotted jaws and no white finger pads |
exo_camera_lowered |
true for the 4,112 episodes recorded after the top camera was mounted lower on 2026-09-28, false for the rest |
exo_camera_height_cm |
height of the top camera's lens above the table in cm, 55 for the 4,112 episodes with the lower camera and 100 for the rest |
exo_view |
the top camera's view, described under Cameras |
exo_camera_world |
the top camera's video in the recording's world frame (see Features): R, a 3 x 3 matrix given as a list of rows, whose columns are the camera's x (right), y (down) and z (forward) axes in world coordinates; t, its position in m, which is (0, 0, 0) where the top camera stood for most of the recording and otherwise where it stood during that episode; and K, the fx, fy, cx, cy of the 784 x 504 video as a pinhole, with pixel centres at integer coordinates. A world point X is at camera coordinates Rᵀ(X − t). null for a recording that has no world frame, and for an episode during which the top camera was moved |
Features
Actions are recovered from the wrist cameras. Each wrist camera is rigidly mounted to its gripper, so the camera's pose
gives the gripper's pose, and each action is the pose change between consecutive frames. Poses come from visual SLAM,
are scaled to metric units with the gripper's IMU, and are checked against the IMU on every frame. Every pose,
action and opening value has a validity flag, and a value whose flag is false is NaN, so it cannot be used by accident.
The statistics in meta/stats.json and in each episode's row of meta/episodes/ leave NaN out.
| Feature | Shape | Meaning |
|---|---|---|
observation.images.exo |
(504, 784, 3) | the top camera |
observation.images.left_wrist, observation.images.right_wrist |
(480, 640, 3) | the left and right wrist cameras |
t |
(1,) | the frame's time in seconds from the start of the episode, at 30 fps (equal to timestamp) |
observation.state |
(14,) | left wrist then right wrist, 7 values each: gripper tip position xyz (m), rotation vector (rad), gripper opening (0 closed, 1 fully open) |
observation.state_valid |
(14,) | whether each value of observation.state is valid |
action |
(14,) | left wrist then right wrist, 7 values each: the step from this frame to the next, as translation (m) and rotation vector (rad) in the wrist's own frame at this frame, then the change in opening. NaN in an episode's last frame |
action_valid |
(14,) | whether each value of action is valid; false in an episode's last frame |
left_chunk, right_chunk |
(1,) | the tracking chunk whose world frame the wrist's position is in (-1 in the rare frames no chunk covers) |
left_pose_source, right_pose_source |
(1,) | how the position was obtained (table below) |
left_rot_action_source, right_rot_action_source |
(1,) | 0 rotation from visual SLAM, 1 from the gyroscope, 2 from a rotation interpolated across a gap where no tracked rotation exists, -1 invalid and in an episode's last frame |
left_filled, right_filled |
(1,) | frames whose position was interpolated across a gap of up to 2 s (pose source 5) |
left_gravity_cam, right_gravity_cam |
(3,) | the unit vector of gravity (pointing down) in the wrist camera's frame (x right, y down, z forward) |
left_pose_world, right_pose_world |
(6,) | the gripper tip's pose in the recording's world frame: position xyz (m), then rotation vector (rad), NaN where it is not known |
Each row also has LeRobot's timestamp, frame_index, episode_index, index and task_index, and LeRobotDataset
returns the row's task as task. Every frame of an episode has the episode's task.
Where the states at k and k+1 are both valid and in the same chunk, the action is exactly the change in state. For rotation matrices R and positions p:
- translation: R_kᵀ (p_{k+1} − p_k)
- rotation: the rotation vector of R_kᵀ R_{k+1}
- gripper: g_{k+1} − g_k
Each chunk has its own world frame, so positions from different chunks cannot be compared directly. The local actions are in the wrist's own frame, so they stay valid across a change of chunk. Each step takes its action from a chunk that tracked both of its ends, so an action can be valid where one of the two states is not.
A chunk's world frame is in general not aligned with gravity. The gravity features give the direction of gravity in each
wrist camera's own frame at every frame, so a pose can be put in a gravity-aligned frame. In the chunk's world frame,
gravity at frame k is R_k g_k, where R_k comes from the rotation in observation.state and g_k is the row of the gravity feature. Gravity comes from
the wrist camera's accelerometer, averaged over the chunk in the chunk's world frame. A row is NaN where the wrist's
rotation is not valid, and in chunks where gravity could not be checked, either because it disagrees with the
accelerometer by more than 3 degrees while the wrist moves or because the chunk's tracking could not be reproduced.
Every recording also has one world frame that both wrists and the top camera share. Its origin is the top camera's
lens where it stood for most of the recording, z points up (against gravity), x is horizontal and points the way the
top camera was tilted there, close to the direction of the top of the top camera's image, and y completes a right-handed frame. In some recordings the top camera was knocked or turned; the world frame stays where it was,
and each episode's exo_camera_world is the camera where it stood during that episode.
left_pose_world and right_pose_world give each gripper tip's pose in it, so the two grippers can be compared
directly, across chunks and across all the episodes of one recording. Within a chunk, the world pose is the pose in
observation.state moved by one rigid transform, so it changes exactly as observation.state does, and the local actions hold in either.
Positions in the world frame are accurate to about 3.5 cm for the position of one gripper relative to the other (median on a calibration recording, measured against markers seen by all three cameras). A gripper's absolute position can be off by a few percent of its distance from the top camera, about 2 to 4 cm. A row of left_pose_world or right_pose_world is
NaN where that wrist's position or rotation in observation.state is not valid, and for every frame of a chunk or a wrist that could not be placed in the world frame. Over both wrists, 87.8% of the frames with a valid position and rotation have a world pose. 11,555 of the 12,219 episodes have world poses, and 11,370 have exo_camera_world.
exo_camera_world in meta/episode_info.jsonl places the top camera's video in the same frame. Because the video is a flat view rendered from a fisheye lens, its pinhole K places a point a few pixels from where the video itself shows it (the camera model alone, apart from the accuracy of the poses above): about 4.3 px median, and 4.7 px for the views of the lowered camera.
| Pose source | Meaning |
|---|---|
| 0 | tracked by visual SLAM |
| 1 | joined from a second SLAM track of the same stretch |
| 2 | bridged with the IMU across a short gap |
| 4 | bridged with the IMU across a longer gap, and confirmed by the top camera |
| 5 | interpolated across a gap of up to 2 s |
| -1 | no valid position (the rotation can still be valid) |
Code 3 is not used.
Share of valid steps, over both wrists and every episode:
| Translation | 91.8% |
| Rotation | 99.98% |
| Gripper opening | 98.7% |
| Both wrists translated in the same step | 85.2% |
The gripper opening is measured from the wrist camera, from fully closed (0) to fully open (1). Operators rarely open the jaws all the way: the median episode's widest opening is 0.68, and 18% of episodes reach 0.95.
Labels
The labels of every episode are in their own dataset, DiverseUMI-sample100-annotations, one line per
episode with its episode_id, which meta/episode_info.jsonl gives for every episode_index. Its card describes every
field.
People in the videos
The wrist and top cameras show the operators, and sometimes people passing by. Faces are not blurred. The license forbids trying to identify anyone, extracting biometric data, and publishing images in which a person could be identified (Article 7).
Loading an episode
import json
from huggingface_hub import hf_hub_download
from lerobot.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("Pantheon-inc/DiverseUMI-sample100")
frame = ds[0]
print(frame["task"], frame["observation.images.exo"].shape) # (3, 504, 784), values from 0 to 1
if frame["action_valid"][:3].all():
left_translation = frame["action"][:3] # m to the next frame, in the left wrist's frame
episodes = [json.loads(l) for l in open(ds.root / "meta/episode_info.jsonl")]
labels = {r["episode_id"]: r["labels"] for r in map(json.loads, open(hf_hub_download("Pantheon-inc/DiverseUMI-sample100-annotations", "labels.jsonl", repo_type="dataset")))}
ep = episodes[int(frame["episode_index"])]
print(ep["episode_id"], ep["exo_camera_world"], labels[ep["episode_id"]]["completion"]["outcome"])
- Downloads last month
- 1,366