Abstract:Unsupervised non-rigid point cloud shape correspondence underpins a multitude of 3D vision tasks, yet itself is non-trivial given the exponential complexity stemming from inter-point degree-of-freedom, i.e., pose transformations. Based on the assumption of local rigidity, one solution for reducing complexity is to decompose the overall shape into independent local regions using Local Reference Frames (LRFs) that are invariant to SE(3) transformations. However, the focus solely on local structure neglects global geometric contexts, resulting in less distinctive LRFs that lack crucial semantic information necessary for effective matching. Furthermore, such complexity introduces out-of-distribution geometric contexts during inference, thus complicating generalization. To this end, we introduce 1) EquiShape, a novel structure tailored to learn pair-wise LRFs with global structural cues for both spatial and semantic consistency, and 2) LRF-Refine, an optimization strategy generally applicable to LRF-based methods, aimed at addressing the generalization challenges. Specifically, for EquiShape, we employ cross-talk within separate equivariant graph neural networks (Cross-GVP) to build long-range dependencies to compensate for the lack of semantic information in local structure modeling, deducing pair-wise independent SE(3)-equivariant LRF vectors for each point. For LRF-Refine, the optimization adjusts LRFs within specific contexts and knowledge, enhancing the geometric and semantic generalizability of point features. Our overall framework surpasses the state-of-the-art methods by a large margin on three benchmarks. Code and models will be publicly available.

A Local-Global Commutative Preserving Functional Map for Shape Correspondence.

3D Shape Matching Using Collinearity Constraint

Robust Structure-based Shape Correspondence

Deformable shape matching with multiple complex spectral filter operator preservation

A Shape Matching Framework Using Metric Partition Constraint

Local Affine Preservation with Motion Consistency for Feature Matching of Remote Sensing Images

Exploiting Local Linear Geometric Structure for Identifying Correct Matches

Geometrically Consistent Partial Shape Matching

Multi-part shape matching by simultaneous partial functional correspondence

Equivariant Local Reference Frames for Unsupervised Non-rigid Point Cloud Shape Correspondence

Integrating Efficient Optimal Transport and Functional Maps For Unsupervised Shape Correspondence Learning

Efficient and Robust Shape Correspondence via Sparsity-Enforced Quadratic Assignment

MCM: an Efficient Geometric Constraint Method for Robust Local Feature Matching.

Locality preserving verification for image search.

Local Geometric Consistency Constraint For Image Retrieval

Efficient Deformable Shape Correspondence Via Multiscale Spectral Manifold Wavelets Preservation

Contour Based Shape Matching For Object Recognition

Feature Matching Using Quasi-Conformal Maps.

Image Matching Based Local Delaunay Triangulation and Affine Invariant Geometric Constraint

Robust Image Matching Via Local Graph Structure Consensus

Guided Locality Preserving Feature Matching for Remote Sensing Image Registration.