Apache 2.0 (Hub tag); MIT (GitHub README)
The publisher's licence permits commercial use and redistribution, with attribution. The Hub card tag is apache-2.0 and the GitHub README says the dataset is released under the MIT License; the two sources name different permissive licences and both permit commercial use with attribution. Confirm which applies before redistribution.
hf download USC-PSI-Lab/humanoid-everyday --repo-type dataset --local-dir ./humanoid-everydayThe older id USC-GVL/humanoid-everyday redirects here. Raw per-task zips (about 500 GB) are linked from a Google Sheet in the GitHub README; scripts/he2lerobot.py converts them. The Hub viewer currently errors on one zero-byte Parquet file.
Specification, as published
- Hours (derived)
- 28 h
- Episodes (stated)
- 10,300 episodes
- Scale
- 10.3k trajectories, 260 tasks x 40 episodes
- Task
- Pick & place, Tidying & put-away, Laundry & folding, Navigation & walking, Other
- Environment
- Indoor – home, Indoor – office, Indoor – lab
- Modality
- RGB, Depth, LiDAR, Tactile, IMU, Proprioception, Actions, Language
- Capture
- Egocentric
- In frame
- Robot hands
- Captured in
- United States of America
- Embodiment
- Unitree G1 with Dex3-1 hands, Unitree H1 with Inspire hands
- Frame rate
- 30 fps
- Resolution
- 640x480
- Formats
- LeRobot v2.0, PNG, PCD, zip
- Size
- 935 GB
- Hosted on
- Hugging Face
Figures are the publisher's own. Kinetic Blocks has not measured this dataset. Derived from the paper's over 3 million frames at 30 Hz; a lower bound, the publisher states no hours figure.
About this dataset
Humanoid Everyday records Unitree G1 (Dex3-1 hands) and H1 (Inspire hands) humanoids teleoperated with Apple Vision Pro across 260 tasks in seven categories: basic, deformable, articulated, tool use, high-precision, human-robot interaction and loco-manipulation, in open-world everyday settings. Streams at 30 Hz: egocentric RGB 640x480, depth, Livox LiDAR point clouds, tactile on the G1, IMU, odometry, arm, leg and hand joint states and actions, the operator's Vision Pro hand and head actions, and language annotations. Distributed in LeRobot v2.0 on the Hub with raw per-task zips (PNG colour, uint16 depth, PCD LiDAR) linked from the GitHub README. The hours figure is a lower bound derived from the paper's over 3 million frames at 30 Hz.
Publisher's own task names: 260 tasks in 7 categories, tool use, articulated objects, loco-manipulation, human-robot interaction.
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.