Datasets that feed robot vision and SLAM.
Large-scale benchmark and baseline for general object grasping.
3,600 hours of egocentric human video used to pretrain manipulation models.
Analytic grasp datasets and the GQ-CNN grasp quality network.
