ASM-6D: Fast Certified Pose-Shape Inference with Complete Correspondence Flow Maps

ASM-6D is a research paper that presents a method for category-level 6D pose and shape estimation using generative complete-geometry prediction and explicit pose-shape inference through shared correspondences. Given an RGB-D image, the method generates alternative complete 3D correspondence hypotheses that are fitted directly for rotation, translation, metric scale, and instance shape. It achieves state-of-the-art performance on ROPE and HouseCat6D benchmarks and supports applications such as shape completion, pose ambiguity, tracking, and robotic grasping at 6-38 Hz.