SAMPLEDataset viewer

A delivered unit,
exactly as it ships.

One folding task per unit: both ego views on one clock, the instruction a reviewer approved, the action steps with what each hand is doing, the moments where the target and the grasp point are grounded on the frame, and the recorded end-effector state underneath. Nothing here is a mock-up of the format.

Units in this batch
523
Action steps
2,555
Distinct targets
188
cam L
cam R · same clock
left hand

right hand

next action

0.0 / 0.0 s
Task

The instruction, the action steps and the grounding are all reviewed by a person before the unit ships. Anchors are drawn only at the moments they were produced for, never interpolated between them.

Gripper

Jaw angle as recorded. Vertical marks are the grounded anchors.

End-effector path

Both hands from above, in metres, cursor at the current moment.

7-DoF end-effector pose · recorded, both hands