Datasets that feed robot vision and SLAM.
100 hours of first-person kitchen video with action and object labels.
