RoboPose

Robot manipulation recordings for moving a cup, stacking cups, and tossing, provided in UMI and teleoperation variants. The original archives combine RGB observations, end-effector positions and axis-angle orientations, and gripper signals. Teleop files additionally contain seven-joint robot state arrays.

Files are organized by task and collection method. Browse real RGB observations, end-effector poses, gripper signals, and episode indices in Data Studio, or download and read the original Zarr archives directly.

Browse real frames in Data Studio · Full episode index · Task catalog · Original files

Full source dataset: 5 archives · 1,569 stored episodes · 451,505 timesteps · 106.38 GB.

Data Studio inspection layer: 120 sampled frames from 15 episodes. The image table is a preview; it is not the complete training dataset. The episode index covers every stored episode.

Explore the data

Data Studio subset	Rows	What you can inspect
Real frame preview	120	Images, task descriptions, episode/frame IDs and original numeric signals
Complete episode index	1,569	Every episode's source archive, start/end frame, length and available FPS
Task / archive catalog	5	A representative observation and full counts for each original archive

Use the subset selector in Data Studio, then filter by task, collection_mode, or episode_id. Image cells contain actual stored pixels. Position, gripper and force columns use the corresponding original frame values. Source archive paths and SHA-256 hashes are included for traceability.

Episode observations

Eight evenly spaced observations from selected original episodes. Animation speed is illustrative, not the original recording rate. UMI tiles show two stored RGB views; teleop files have one RGB view.

Full-dataset coverage

Counts are computed from the original arrays, independently of the small frame preview. The five archives cover three tasks; UMI and teleop variants remain separate.

An episode's numeric signals

The plot uses eight real observations from one episode. Lines connect sampled values for visualization; coordinates and signals are shown in their stored conventions.

Observation gallery across every original archive

Original archives

File	Collection	Episodes	Timesteps	Size (GB)
moving_cup_UMI.zarr	UMI	385	48,246	14.53
moving_cups_teleop.zarr	teleop	182	58,329	8.81
stack_cups_UMI.zarr	UMI	400	139,022	41.87
stack_cups_teleop.zarr	teleop	180	138,861	20.96
toss_UMI.zip	UMI	422	67,047	20.20
Total		1,569	451,505	106.38

Sizes are rounded decimal GB. Exact sizes, source hashes, array schemas and sample indices are recorded in showcase_manifest.json. Statistics were verified from the original archives on 2026-10-27.

Modalities and fields

Modality	Original array	Shape / availability
RGB	data/camera0_rgb	(T, 224, 224, 3), all files
Second RGB view	data/camera1_rgb	(T, 224, 224, 3), UMI
End-effector position	data/robot0_eef_pos	(T, 3)
Axis-angle orientation	data/robot0_eef_rot_axis_angle	(T, 3)
Gripper signal	data/robot0_gripper_width	(T, 1)
Joint position / velocity / external torque	robot0_joint_pos, robot0_joint_vel, robot0_joint_tau_ext under data/	(T, 7), teleop
Episode boundaries	meta/episode_ends	(E,), cumulative exclusive ends

UMI files also store head-relative end-effector fields separately. Teleop files contain TCP/flange poses, measured gripper width, gripper force and timestamps. Both teleop files report 30 FPS and binary-gripper metadata; UMI FPS is not specified in the inspected metadata. Consult the original collection/export pipeline for units and coordinate-frame conventions.

Download and load

The preview tables can be loaded directly with the Hugging Face Datasets library.

pip install -U huggingface_hub datasets "zarr<3"

Load the small inspection tables

from datasets import load_dataset

frames = load_dataset("lihong-cs/robopose", "frame_preview", split="preview", streaming=True)
sample = next(iter(frames))
print(sample["task"], sample["episode_id"], sample["frame_index"])
sample["image"].save("preview-rgb.png")

episodes = load_dataset("lihong-cs/robopose", "episode_index", split="episodes", streaming=True)
print(next(iter(episodes)))

Read an original Zarr archive

Download one complete source archive:

hf download lihong-cs/robopose moving_cup/moving_cup_UMI.zarr --repo-type dataset --local-dir ./data

import zarr

path = "./data/moving_cup/moving_cup_UMI.zarr"
with zarr.ZipStore(path, mode="r") as store:
    root = zarr.open_group(store=store, mode="r")
    episode_ends = root["meta/episode_ends"][:]
    episode_index = 0
    start = 0 if episode_index == 0 else int(episode_ends[episode_index - 1])
    stop = int(episode_ends[episode_index])
    position = root["data/robot0_eef_pos"][start:stop]
    orientation = root["data/robot0_eef_rot_axis_angle"][start:stop]
    gripper = root["data/robot0_gripper_width"][start:stop]
    rgb = root["data/camera0_rgb"][start]
    print(position.shape, orientation.shape, gripper.shape, rgb.shape)

Read image arrays by individual frame or short slice. Downloading an archive retrieves the full original file; loading frame_preview retrieves the small inspection subset. Use episode-level splits for training and evaluation.

Preview construction and provenance

• The complete episode index comes from every meta/episode_ends array. Counts and array lengths are checked for consistency.
• Frame previews sample the first, middle and final stored episode in each archive, with eight evenly spaced timesteps per selected episode.
• Each preview image is stored losslessly in Parquet with a typed Hugging Face Image feature. Frame indices and source hashes identify the corresponding original data.
• Preview images and plots are additional inspection assets. Original Zarr archive contents are unchanged.
• time_in_episode_s_nominal and duration_s_nominal are derived only where FPS metadata exists. They are not replacement source timestamps.

Original source: ModelScope: lihongcs/robopose.

License

This dataset is licensed under CC BY 4.0. Commercial use is permitted under the license terms, including appropriate attribution, a link to the license, and an indication of any changes.