Concentric Spherical Neural Network for 3D Representation Learning

Authors

James Fox, Georgia Institute of Technology
Bo Zhao, University of California, San Diego
Beatriz Gonzalez Del Rio, Universidad de Valladolid
Sivasankaran Rajamanickam, Sandia National Laboratories
Rampi Ramprasad, Georgia Institute of Technology
Le Song, Mohamed Bin Zayed University of Artificial IntelligenceFollow

Document Type

Conference Proceeding

Publication Title

Proceedings of the International Joint Conference on Neural Networks

Abstract

Learning 3D representations of point clouds that generalize well to arbitrary orientations is a challenge of practical importance in domains ranging from computer vision to molecular modeling. The proposed approach uses a concentric spherical spatial representation, formed by nesting spheres discretized the icosahedral grid, as the basis for structured learning over point clouds. We propose rotationally equivariant convolutions for learning over the concentric spherical grid, which are incorporated into a novel architecture for representation learning that is robust to general rotations in 3D. We demonstrate the effectiveness and extensibility of our approach to problems in different domains, such as 3D shape recognition and predicting fundamental properties of molecular systems.

DOI

10.1109/IJCNN55064.2022.9892358

Publication Date

9-30-2022

Keywords

Representation learning, Point cloud compression, Solid modeling, Computer vision, Three-dimensional displays, Shape, Computational modeling

Comments

IR conditions: non-described

Recommended Citation

J. Fox, B. Zhao, B. G. del Rio, S. Rajamanickam, R. Ramprasad and L. Song, "Concentric Spherical Neural Network for 3D Representation Learning," 2022 International Joint Conference on Neural Networks (IJCNN), 2022, pp. 1-8, doi: 10.1109/IJCNN55064.2022.9892358.