ASM-6D: Fast Certified Pose-Shape Inference with Complete Correspondence Flow Maps
We present ASM-6D, which couples generative complete-geometry prediction with explicit category-level pose-shape inference through shared correspondences.
ProofPaper ↗
Key points
- Category-level 6D pose and shape estimation requires recovering an object's pose and geometry from partial observations without an exact instance model.
- Given an RGB-D image, ASM-6D generates alternative complete 3D correspondence hypotheses whose point identities are shared with a category-level Active Shape Model (ASM), including occluded points.
- We analytically eliminate translation and shape coefficients, reducing fitting to a nine-variable scaled-rotation optimization problem that enables efficient batched evaluation of multiple hypotheses in real time.
- Ablations demonstrate the importance of generative correspondence prediction and active shape variation.
Sources (1)
- [1]ASM-6D: Fast Certified Pose-Shape Inference with Complete Correspondence Flow MapsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 02:41 AM
We present ASM-6D, which couples generative complete-geometry prediction with explicit category-level pose-shape inference through shared correspondences.
Category-level 6D pose and shape estimation requires recovering an object's pose and geometry from partial observations without an exact instance model.
Extractive summary: sentences quoted from the sources.