Robot learning dataset
Humanoid Everyday
Open-world humanoid manipulation dataset of about 10,300 trajectories and over 3 million frames across 260 tasks, recorded on Unitree G1 and H1 with RGB-D, LiDAR, tactile and language annotations. The project page shows no download link and no dataset license.
Records last verified Oct 13, 2026 · Send a correction
Licence & use
Licence not established
Commercial use: not established
Terms apply to the checked release. Read publisher conditions for mixed collections, converted formats and gated access.
- Publisher
- University of Southern California / Toyota Research Institute
- Released
- Not dated
- Published size
- 10,300 trajectories
- Modalities
- rgb, depth, lidar, tactile, imu, proprioception, language
- Formats
- Not disclosed
- Real / simulated
- Real
- Includes humanoid recordings
- Yes
Size & release notes
Project and paper publish 10.3k trajectories, over 3 million frames and 260 tasks across seven categories. The project page provides no dataset download or license, despite the paper stating that they are publicly available; its evaluation link is still marked Coming Soon. Paper publication is not a verified data-release date.
Recorded embodiments
More robot learning datasets
Evidence
Sources
- [s1] Humanoid Everyday project page ↗
Humanoid Everyday authors · official · accessed Oct 13, 2026
- [s2] Humanoid Everyday: A Comprehensive Robotic Dataset for Open-World Humanoid Manipulation ↗
arXiv · paper · published Oct 9, 2025 · accessed Oct 13, 2026
- [s3] Humanoid Everyday paper: platforms and multimodal collection ↗
arXiv · paper · published Jul 4, 2026 · accessed Oct 13, 2026