Dense Shape Correspondence

Dense shape correspondence aims to establish point-wise mappings between the surfaces of 3D shapes, even across variations in topology or geometry. Recent research focuses on developing deep learning models, including implicit functions, graph convolutional networks, and neural radiance fields (NeRFs), to learn these correspondences in an unsupervised or weakly supervised manner. These methods leverage geometric features and learned embeddings to achieve robust and accurate correspondences, improving upon traditional approaches. The resulting advancements have significant implications for various applications, such as shape analysis, object recognition, and image-based tasks like texture transfer and annotation.

Papers

December 29, 2022

Learning Implicit Functions for Dense 3D Shape Correspondence of Generic Objects
Feng Liu, Xiaoming Liu
Dense Correspondence Semantic Correspondence Generic Model Shape Latent Deep Implicit Function Dense Shape Correspondence

October 17, 2022

Anisotropic Multi-Scale Graph Convolutional Network for Dense Shape Correspondence
Mohammad Farazi, Wenhui Zhu, Zhangsihao Yang, Yalin Wang
Deep Convolutional Neural Network Multi Scale Graph Geometric Deep Learning Model Discretization Invariant Dense Shape Correspondence

November 8, 2021

Template NeRF: Towards Modeling Dense Shape Correspondences from Category-Specific Object Images
Jianfei Guo, Zhiyuan Yang, Xi Lin, Qingfu Zhang
Neural Radiance Field Dense Correspondence Implicit Template Dense Shape Correspondence

Dense Shape Correspondence

Papers

Learning Implicit Functions for Dense 3D Shape Correspondence of Generic Objects

Anisotropic Multi-Scale Graph Convolutional Network for Dense Shape Correspondence

Template NeRF: Towards Modeling Dense Shape Correspondences from Category-Specific Object Images