kt_pick_pringles_spam_merge

Two-task dataset: merge of learner1119/kt_pick_pringles_merge (merged index 0–99) and learner1119/kt_pick_spam_merge (merged index 100–199). FFW-SH5 rev1, 10 fps, LeRobot v2.1.

task_index	language instruction	episodes	frames	source
0	pick up the pringles can	100 (0–99)	7,195	kt_pick_pringles + kt_pick_pringles2
1	pick up the spam	100 (100–199)	9,021	kt_pick_spam + kt_pick_spam2
	total	200	16,216

Episodes were re-indexed end to end; videos are copied byte-for-byte. meta/source_episodes.jsonl maps each episode back to the per-task merge it came from.

Features

• observation.state, action: 57-dof (arm_l 0:7, arm_r 7:14, finger_l 14:34, finger_r 34:54, head 54:56, lift 56)
• observation.tactile: 90
• observation.images.cam_head (672×376), observation.images.cam_head_realsense (640×480)

meta/modality.json (GR00T)

Exposes only the right arm (7) + right hand (20) as right_arm / right_hand and only cam_head as video. The other joints and the realsense camera stay in the data but are not used for training.

Hold-out

2 episodes from each of the 4 recording cases are held out (8 episodes, 580 frames), published as learner1119/kt_pick_pringles_spam_merge_holdout:

case	source	episodes here
Pringles, right of Spam	kt_pick_pringles	19, 39
Pringles, left of Spam	kt_pick_pringles2	69, 89
Spam, set 1	kt_pick_spam	119, 139
Spam, set 2	kt_pick_spam2	169, 189

The remaining 192 episodes (15,636 frames, 96 per task) are the training split. The Spam part equals learner1119/kt_pick_spam_merge_holdout; the Pringles part is a subset of learner1119/kt_pick_pringles_merge_holdout.