Dataset · airoa-org
YUBI UMI Arena CoRL 2026 video dataset part 5
airoa-org/yubi-corl2026-umi-arena-05-3k2
Part 5 of the massive AIRoA stereo video corpus for robot learning from the UMI Arena competition at CoRL 2026.
What's inside
This is the final, fifth part of the complete YUBI data release from the AI Robot Association (AIRoA), prepared for The UMI Arena competition at CoRL 2026. The full release includes over 1.23 million episodes, more than 350 million frames, and 3,242.3 hours of 30 Hz recordings across 5,849 tasks. Data is provided as dual-view video streams (left and right cameras) with English descriptions and is distributed under the Limited-Access Dataset Terms of Use.
What you can do with it
- Training Vision-Language-Action (VLA) robot control models and imitation learning policies.
- Training visual encoders for stereo vision and spatial manipulation understanding.
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 22K
- Likes
- 0
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("airoa-org/yubi-corl2026-umi-arena-05-3k2")