Morning Edition · Saturday, October 24, 2026Published at 2:20 AM EDT · New York

Fudan's F2G-Pose predicts 6D pose, metric size and full shape from one RGB-D frame at 28 frames per second

The NeurIPS 2026 method lifts foundation-model features onto point clouds and routes them through an eight-expert transformer, running about seven times faster than diffusion-based GenPose++.

F2G-Pose, from researchers at Fudan University, is a category-level object pose estimator. From a single segmented RGB-D observation it predicts 6D pose, metric size and a completed camera-frame shape in one feed-forward pass. The code is r…