First UMI dataset collected: faster capture than tele-op, but mimicking robots is exhausting
DominiqueCAPaul · x · 2026-09-16
ginokreis reports collecting the first UMI dataset, with humans imitating robot motions for a meaningful comparison against teleoperation data.
Key takeaways: collection and scene-reset are much faster than tele-op; imitating the robot was exhausting for the human demonstrator; and UMI needs more control and feedback integrated (beyond just a Mac and keyboard). A hands-on look at embodied data collection workflows.
More from Embodied
- Odyssey unveils Odyssey-3 foundation world model for robots, humanoids and driving — chris_j_paxton · 2026-09-16
- The real bottleneck in humanoid robots is data, not hardware — ingliguori · 2026-09-16
- Attach Claude, Codex and Gemini agents to RealSense cameras with one npx command — chrismatthieu · 2026-09-16
- One researcher taught world model Odyssey-3 to drive in India with just 20 hours of data — soleio · 2026-09-16
- Throwing 3D models from a Meta Quest straight into a 3D printer — OpenAIDevs · 2026-09-16
- Wayve Co-founder Explains Its Contrarian End-to-End Embodied AI Bets at Actuate26 — Sethwinterroth · 2026-09-16