Fudan's F2G-Pose directly regresses 6D pose from one RGB-D frame, 7x faster
机器之心 · wechat · 2026-11-22
F2G-Pose (NeurIPS 2026, Fudan + Shanghai Innovation Institute) takes a direct-regression route to category-level 6D pose estimation: one forward pass over a segmented RGB-D frame predicts 6D pose, true size, and camera-frame shape completion — no diffusion sampling or candidate filtering. It lifts frozen RADIOv2.5 foundation features onto point clouds via camera intrinsics, encodes local geometry with DGCNN, aggregates with an MoE Transformer (8 experts, top-2), and uses shape completion as geometric supervision. The paper also exposes GenPose++'s selection bottleneck: 79.3% of instances have a <5° rotation candidate, but only 48.3% survive filtering.
More from Embodied
- Kyros Labs builds Kepler, one workspace to replace 5-tool robot policy debugging — Repulsive-Fee-2735 · 2026-11-22
- Musk says Optimus will wash dishes and take out trash within 12-18 months — nikola_mr64990 · 2026-11-22
- Driving 96.6% of miles on FSD saves just $3.32 on Tesla insurance — altryne · 2026-11-22
- Open-source visloc-rs turns 360° camera street footage into a robot simulator via 3D Gaussian Splatting — rsasaki0109 · 2026-11-22
- Blogger asks Tesla FSD V15 demo posters to show the outside view for verification — justin_hart · 2026-11-22
- iPhone 20 Pro rumored to jump to 200MP main sensor with pixel binning — aftahi_ai · 2026-11-22