Datasets pooled across many robots.
15 million video frames from 7 robot platforms for large-scale visual learning.
