Datasets:

jezv
/

takeitback-tray-pd-joint-pos

Tasks:

Reinforcement Learning

Modalities:

Languages:

Size:

Libraries:

Dataset card Data Studio Files Files and versions

Episodes Preview ds_fetch Visualizer Open in visualizer

1.21k episodes · 10 fps · 3 cameras · 256×256 mp4v

MIKASA TakeItBack Tray — DSFetch

Episodes: 1206 successful trajectories
Frequency: 10 Hz; source: pd_joint_pos at 20 Hz, stride 2
Action shape: 13
Policy state: 12D (qpos[3:]); global_state stores base pose separately
RGB cameras: fetch_hand 128x128, left/right base cameras 256x256
Source seeds: 1700–3420; 1206 successful episodes
Waypoint jitter: seeded uniform +/-5% of local waypoint scale
Robot: project ds_fetch; unsuccessful attempts omitted

Language instruction: Place the cup onto the tray.

Downloads last month: 65

Number of rows:

Total file size: