Home · Open data · Manipulation · Human video

EgoDemo (EgoSuite-Open100K sample)

50 hours: a sample of Lightwheel's EgoSuite-Open100K, first-person stereo video of everyday work tasks with 3D hand and body pose, in LeRobot v3 and MP4.

Published by Lightwheel AI, 2026-08

Open data
Training OK, no resale

Commercial Training, No Resale v1.0 (Lightwheel)

PriceFree

The publisher's licence permits commercial training but not redistribution or resale of the data. The Hub tag reads commercial-training-no-resale-v1.0 and the publisher's release note says academic research and commercial training are permitted; resale of the data is not. The licence document itself is behind the access form.

Request access at source Opens huggingface.co in a new tab

The publisher approves access by hand. Expect to fill in a short form and wait.

From a terminal
hf download LightwheelAI/EgoDemo --repo-type dataset --local-dir ./EgoDemo

Accept the access terms on the Hugging Face page (name, affiliation, intended use), then sign in with the Hugging Face CLI before downloading. The full EgoStandard and EgoPro sets use a separate bucket sync described on the publisher's pages.

Specification, as published

Hours (stated)
50 h
Scale
a 50-hour sample spanning all four SKUs of a 100,000-hour corpus
Task
Pick & place, Sorting & packing, Assembly, Cooking & food prep, Repair & maintenance, Tidying & put-away +1 more
Environment
Indoor – home, Indoor – office, Indoor – warehouse, Indoor – studio
Modality
RGB, Stereo, Hand pose, Language
Capture
Egocentric
In frame
Person hands
Captured in
Multiple regions
Embodiment
human
Formats
MP4, LeRobot v3, MCAP, Parquet
Size
1.48 TB
Hosted on
Hugging Face

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

About this dataset

EgoDemo is the openly downloadable sample of EgoSuite-Open100K, Lightwheel AI's egocentric human activity corpus collected by a distributed workforce across home, hospitality, retail, sports, logistics, office and industrial scenes. The sample spans all four subsets: EgoStand and EgoStand-Body (head-mounted stereo cameras, 3D hand pose, and full-body pose in the Body variant) in LeRobot v3 and MCAP form with temporal event annotations, plus two raw-video variants of undistorted MP4 with camera parameters and no annotations. Recordings were de-identified before release and participants gave informed consent. Access is by a short form on Hugging Face; files download with the Hugging Face CLI once approved.

Publisher's own task names: assembly, cooking, inventory management, tool use, repair, packing, Affix Shipping Label (on Package Surface), Align Snack Bags to Shelf Front Edge, Apply Takeout Cup Label.

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.