Learning Spectral Transform Network on 3D Surface for Non-rigid Shape Analysis
This work addresses the challenge of designing networks for non-rigid shape analysis in 3D computer vision, with incremental improvements in specific benchmarks.
The paper tackled the problem of non-rigid shape analysis on 3D surfaces by proposing a spectral transform network, achieving the highest accuracies on SHREC14, SHREC15, and the Range subset of SHREC17 datasets for shape retrieval and classification.
Designing a network on 3D surface for non-rigid shape analysis is a challenging task. In this work, we propose a novel spectral transform network on 3D surface to learn shape descriptors. The proposed network architecture consists of four stages: raw descriptor extraction, surface second-order pooling, mixture of power function-based spectral transform, and metric learning. The proposed network is simple and shallow. Quantitative experiments on challenging benchmarks show its effectiveness for non-rigid shape retrieval and classification, e.g., it achieved the highest accuracies on SHREC14, 15 datasets as well as the Range subset of SHREC17 dataset.