Home · Open data · Manipulation · Robot video

OpenNeoData

5,000+ hours and 200k+ trajectories of visuotactile manipulation across ALOHA, ARX-5, Flexiv, UR and handheld UMI embodiments, CC BY-NC-SA 4.0.

Published by NeoteAI with Fudan University TEAI, 2026-07

Open data
Non-commercial only

CC BY-NC-SA 4.0

PriceFree

The publisher's licence restricts use to research and other non-commercial purposes. Card states: All the data and code within this repo are under CC BY-NC-SA 4.0. Non-commercial.

Get dataset Opens huggingface.co in a new tab

The host asks you to sign in to Hugging Face and accept the publisher's terms before it serves files.

From a terminal
hf download NeoteAIEmbodied/OpenNeoData --repo-type dataset --include ur/** --local-dir OpenNeoData

Contact-information gate; sign in first. The card documents per-embodiment pulls (ur, aloha, arx and so on) with --include. Mirror on ModelScope as NeoteAI/OpenNeoData.

Specification, as published

Hours (stated)
5,000 h
Episodes (stated)
200,000 episodes
Scale
5,000+ hours, 200k+ trajectories, 257 task types
Task
Pick & place, Assembly, Laundry & folding, Other
Environment
Indoor – lab
Modality
RGB, Tactile, Proprioception, Actions, End-effector pose
Capture
Egocentric
In frame
Robot hands, Person hands
Captured in
Asia
Embodiment
ALOHA dual-arm, ARX-5, Flexiv Rizon 4s, UR5e +2 more
Frame rate
30 fps
Resolution
640x360
Formats
LeRobot v3.0, Parquet, MP4
Size
9.07 TB
Hosted on
Hugging Face

Figures are the publisher's own. Kinetic Blocks has not measured this dataset.

About this dataset

OpenNeoData spans six embodiments, five fixed-arm platforms (ALOHA dual, ARX-5 dual and single, Flexiv Rizon 4s, UR5e and UR7e) and two handheld UMI device families (dual and single hand), all with wrist-mounted visuotactile gel sensors. 257 task types cover contact-rich manipulation, deformable objects, precise placement and bimanual coordination. Each embodiment carries three to seven streams: third-view RGB, wrist RGB and gel visuotactile camera video at 30 fps (HEVC MP4, 640x360), observation state, action and end-effector poses on the fixed arms. The publisher states over 5,000 hours and 200k trajectories in LeRobot v3.0, 9.07 TB, behind a contact-information gate. The UMI subsets are first-person human footage; the rest is robot.

Publisher's own task names: 257 task types, contact-rich manipulation, deformable objects, precise placement, bimanual coordination.

What Kinetic Blocks did, and did not do

Indexed and linked from the publisher. Not hosted, verified or graded by Kinetic Blocks. The publisher's terms govern. The licence shown is the one stated on the publisher's page, linked above. Nothing on this page is a claim by Kinetic Blocks.

Want the verified ones too?

Request access to the marketplace, where every paid listing is verified against its files and graded before it goes on sale.