English

Auditing Training-Free 3D Shape Retrieval with Diffused Geodesic Moments

Computer Vision and Pattern Recognition 2026-05-29 v1 Graphics

Abstract

Reported retrieval scores for training-free shape descriptors conflate local signal design, normalization, aggregation, codebook fitting, and metric choices, making isolated component evaluation difficult. This paper reframes descriptor evaluation as a {\em protocol audit}. We introduce Diffused Geodesic Moments (DGM), a seed-conditioned descriptor that computes sparse implicit heat responses, converts them to distance-like fields, and summarizes each vertex by low-order moments across seeds and scales. DGM is used both as a practical non-spectral baseline and as an instrument for isolating protocol effects. On the registered FAUST benchmark split (FAUST-Reg) and the TOSCA shape collection, aggregation-matched experiments show that an independent Geometric Moment Shape Descriptor baseline built on Heat Kernel Signature features (GMSD-HKS) obtains the highest scores in this implementation (0.621/0.8200.621/0.820 and 0.865/0.9630.865/0.963 mean average precision (mAP)/top-1), Wave Kernel Signature (WKS) remains a strong classical signal, and DGM is useful mainly when sparse solves, non-spectral deployment, or symmetry-informative seed frames are priorities. The broader finding is methodological: the input field and aggregation protocol can dominate the moment formula. The paper contributes a reproducible protocol-cascade analysis, a cross-shape alignment diagnostic for functional-map compatibility, and concrete recommendations for designing and reporting training-free shape descriptors.

Keywords

Cite

@article{arxiv.2605.29004,
  title  = {Auditing Training-Free 3D Shape Retrieval with Diffused Geodesic Moments},
  author = {Zhicheng Du and Changyue Liu and Wenji Xi and Zhaotian Xie and Zhuo Deng and Ziheng Zhang and Yang Liu and Lan Ma},
  journal= {arXiv preprint arXiv:2605.29004},
  year   = {2026}
}