The footage robot learning runs on.
Manipulation
Grasping, placing, pouring, folding. Third-person and wrist views with success labels.
Teleoperation
Operator demonstrations on arms and humanoids, with the command stream when you have it.
Egocentric video
Head-mounted footage of everyday tasks in kitchens, warehouses and workshops.
Navigation
Walk-throughs of homes, offices and sites. Doors, stairs, clutter, people.
Failure cases
Drops, slips, collisions and recoveries. The clips most datasets are missing.
Long-horizon tasks
Multi-step chores filmed start to finish, narrated or silent, to your spec.