StereoPolicy learns robot 3D perception from stereo pairs, hits 59% success
TheZachMueller · x · 2026-09-10
Lambda's StereoPolicy gives robots 3D perception without depth sensors, depth maps, point clouds, or LiDAR — learning 3D structure directly from stereo image pairs. Across 5 real-world tabletop tasks it achieved 59% success vs. 42% for RGB, 41% for RGB-D, and 14% for PointNet.
More from Embodied
- Apple's first foldable costs $1,999 and drops Pro features; Samsung mocks 7-year-late arrival — tech__unicorn · 2026-09-10
- NVIDIA details Alpamayo 2 Super, its L4 autonomy model for robotaxis — drmapavone · 2026-09-10
- iPhone Duo hinge may expose an API — devs probe if it's even private — zats · 2026-09-10
- iPhone Duo called the first agentic cyberdeck — and the timing is finally right — michalmalewicz · 2026-09-10
- Apple's fall event: foldable iPhone Duo debut and an always-listening Apple Watch — TechCrunch AI · 2026-09-10
- Tesla's robotaxi moat is the purpose-built Cybercab line, not software — JOBhakdi · 2026-09-10