UMI DAS Handheld Gripper — Kitchen-Counter Pick-and-Place: Drinking Straws, Cooking Tongs and Condiment Jars

A single showcase episode of handheld-gripper tabletop manipulation, converted from raw MCAP into a HuggingFace LeRobot v2.1 dataset and shipped with the complete QC report for the capture session it came from. The episode — 'arrange the silicone strips on the table' — runs 4,197 frames (139.9 s) at 30 fps and is the strongest of the 13 captured: the longest, with the most gripper open/close cycles (13), the greatest end-effector travel (20.8 m) and the largest workspace envelope (1.2 × 1.03 × 0.78 m). Three synchronised fisheye views — cam_ego, cam_left_wrist, cam_right_wrist — all 640×360 H.264, alongside an 8-DoF state vector and 8-DoF action vector per frame. The bundled qc_report/ folder carries the QC verdict as machine-readable JSON, HTML and Markdown, plus a representative frame. QC outcome: 20/20 checks PASS, zero warnings, zero failures — readability, duration bounds, stream presence, per-stream rate, cross-camera frame match, timestamp monotonicity, inter-frame gaps and camera-to-pose sync. The episode was additionally reviewed frame-by-frame and contains no people, hands, faces or screens.