Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
video
video
11
31.6
label
class label
24 classes
020260901-174943
020260901-174943
020260901-174943
120260901-175030
120260901-175030
120260901-175030
220260901-175110
220260901-175110
220260901-175110
320260901-175303
320260901-175303
320260901-175303
420260901-175348
420260901-175348
420260901-175348
520260901-175525
520260901-175525
520260901-175525
620260901-182421
620260901-182421
620260901-182421
720260901-182501
720260901-182501
720260901-182501
820260901-182537
820260901-182537
820260901-182537
920260901-182714
920260901-182714
920260901-182714
1020260901-182810
1020260901-182810
1020260901-182810
1120260901-182845
1120260901-182845
1120260901-182845
1220260901-182924
1220260901-182924
1220260901-182924
1320260901-183002
1320260901-183002
1320260901-183002
1420260901-183055
1420260901-183055
1420260901-183055
1520260901-183143
1520260901-183143
1520260901-183143
1620260901-183246
1620260901-183246
1620260901-183246
1720260901-183330
1720260901-183330
1720260901-183330
1820260901-183507
1820260901-183507
1820260901-183507
1920260901-184147
1920260901-184147
1920260901-184147
2020260901-184226
2020260901-184226
2020260901-184226
2120260901-184301
2120260901-184301
2120260901-184301
2220260901-184342
2220260901-184342
2220260901-184342
2320260901-185102
2320260901-185102
2320260901-185102

UR5e pi0.5 rollouts

Policy rollouts of mahgoobi/ur5e_pi05_10k on a real UR5e with a Robotiq gripper, task "put the cup in the bowl", collected 2026-09-01 from the deployment page in RoboResearch (roboresearch.evaluation.ur5e.ui). 25 episodes at 20 Hz, action chunks of 50 steps with action_steps of them executed per chunk (25 for every episode but the first, which ran 10). 20 successes, 5 failures. Every failure is in a scene with three or more distractors; the clean and one-distractor scenes succeed throughout.

Layout

One directory per episode, named by its start time, in the layout the training recordings were made in:

file content
state.csv one row per control step: timestamp, joint_0..5, tcp_*, gripper_pos, the three camera filenames, and the commanded action_joint_0..5, action_gripper that went to servoJ from that state
camera1/, camera2/, camera3/ side, wrist, base views as timestamp-named 640x480 PNGs, one per row
camera{1,2,3}.mp4 the same frames as h264 at the recorded size, for watching
camera{1,2,3}.train.mp4 the same frames at the 240x320 the training set stores, encoded as the training conversion encodes
meta.json the run's settings, the checkpoint, and the annotation

The training recordings hold observations only; these also hold the commanded target beside the realized state, so the arm's tracking error is measured directly. The first episode, 20260901-151158, was encoded before this layout settled and holds videos only.

Annotation

Episode level: outcome is success or failure, note is one sentence on what happened with step indices, labelled_by is who judged it. Episodes marked claude were judged from their frames after the collection, from contact sheets of the three cameras; the one without it was labelled from the page during the run.

episode rows outcome by note
20260901-151158 500 success page
20260901-174943 344 success claude clean scene; grasp at 48 after a brief close-reopen, placed in the bowl and released at 170, then 8 s of drift
20260901-175030 335 success claude clean scene; grasp at 68, placed in the bowl at 205
20260901-175110 295 success claude pizza on the table; grasp at 59, placed at 162
20260901-175303 400 success claude pizza; grasp at 83, placed at 235
20260901-175348 393 success claude pizza and apple; grasp at 42, placed at 182, then 10 s of drift
20260901-175525 310 success claude pizza, apple, carrot; grasp at 70, placed at 181
20260901-182421 329 success claude clean scene; grasp at 73, placed in the bowl and released at 190
20260901-182501 270 success claude clean scene; grasp at 43, placed at 204
20260901-182537 410 success claude clean scene, long carry; grasp at 85, placed at 330
20260901-182714 221 success claude clean scene, cup beside the bowl; grasp at 74, placed at 177
20260901-182810 286 success claude clean scene; grasp at 80, placed at 213
20260901-182845 244 success claude pizza; grasp at 78, placed at 205
20260901-182924 284 success claude pizza and lemon; grasp at 84, placed at 234
20260901-183002 631 failure claude pizza, apple, lemon; cup knocked onto its side around 270 and never recovered, gripper closes on nothing near the lemon at 540
20260901-183055 506 failure claude pizza, apple, lemon; grasp at 85 then a stall: holds the cup on the table for 20 s without lifting
20260901-183143 387 failure claude pizza, apple, lemon; grasp at 110, lifted, cup dropped on the table beside the bowl around 330, re-grasp attempt near the lemon
20260901-183246 270 success claude pizza, apple, lemon; grasp at 78, placed at 206
20260901-183330 220 success claude pizza, apple, lemon; grasp at 56, placed at 169
20260901-183507 500 failure claude heavy clutter; first grasp released without lifting at 285, re-grasp at 356, released at 436 leaning on the bowl's rim rather than inside it
20260901-184147 228 success claude clean scene; grasp at 52, placed in the bowl and released at 177
20260901-184226 267 success claude pizza and apple; grasp at 60, placed at 191
20260901-184301 363 failure claude pizza, apple, carrot, lemon; grasp at 88, lifted, then holds the cup tilted against the bowl's rim from about 200 to the end without releasing
20260901-184342 458 success claude pizza, apple, carrot, lemon; grasp at 57, placed in the bowl by 261, then re-grasps the placed cup at 326 and releases it at 342 still in the bowl
20260901-185102 342 success claude pizza, apple, carrot, lemon; a hand moves the cup in view around rows 48 to 146, then grasp at 104, placed at 282; the hand is in the model's observations

Rows recorded after the release are the policy drifting past the end of its trained behaviour; the training recordings end at release. One step gap of about 230 ms every action_steps rows is inference at each chunk boundary.

Previews

Side, wrist and base views of every episode, left to right, at 5 fps; previews/<episode>.gif. The wrist and base views are what the model saw. The scene column is the clutter on the table.

Successes

episode scene side · wrist · base
20260901-151158 clean scene 20260901-151158
20260901-174943 clean scene 20260901-174943
20260901-175030 clean scene 20260901-175030
20260901-175110 pizza on the table 20260901-175110
20260901-175303 pizza 20260901-175303
20260901-175348 pizza and apple 20260901-175348
20260901-175525 pizza, apple, carrot 20260901-175525
20260901-182421 clean scene 20260901-182421
20260901-182501 clean scene 20260901-182501
20260901-182537 clean scene, long carry 20260901-182537
20260901-182714 clean scene, cup beside the bowl 20260901-182714
20260901-182810 clean scene 20260901-182810
20260901-182845 pizza 20260901-182845
20260901-182924 pizza and lemon 20260901-182924
20260901-183246 pizza, apple, lemon 20260901-183246
20260901-183330 pizza, apple, lemon 20260901-183330
20260901-184147 clean scene 20260901-184147
20260901-184226 pizza and apple 20260901-184226
20260901-184342 pizza, apple, carrot, lemon 20260901-184342
20260901-185102 pizza, apple, carrot, lemon 20260901-185102

Failures

episode scene side · wrist · base
20260901-183002 pizza, apple, lemon 20260901-183002
20260901-183055 pizza, apple, lemon 20260901-183055
20260901-183143 pizza, apple, lemon 20260901-183143
20260901-183507 heavy clutter 20260901-183507
20260901-184301 pizza, apple, carrot, lemon 20260901-184301
Downloads last month
247