DeepSphere: a graph-based spherical CNN
arXiv:2012.15000
Abstract
Designing a convolution for a spherical neural network requires a delicate tradeoff between efficiency and rotation equivariance. DeepSphere, a method based on a graph representation of the sampled sphere, strikes a controllable balance between these two desiderata. This contribution is twofold. First, we study both theoretically and empirically how equivariance is affected by the underlying graph with respect to the number of vertices and neighbors. Second, we evaluate DeepSphere on relevant problems. Experiments show state-of-the-art performance and demonstrates the efficiency and flexibility of this formulation. Perhaps surprisingly, comparison with previous work suggests that anisotropic filters might be an unnecessary price to pay. Our code is available at https://github.com/deepsphere
published at ICLR'20, https://openreview.net/forum?id=B1e3OlStPB
References in corpus (7)
- HEALPix -- a Framework for High Resolution Discretization, and Fast Analysis of Data Distributed on the Sphere
- Spectral Networks and Locally Connected Networks on Graphs
- DeepSphere: Efficient spherical Convolutional Neural Network with HEALPix sampling for cosmological applications
- Learning Spherical Convolution for Fast Features from 360° Imagery
- Spherical CNNs on Unstructured Grids
- Cosmological model discrimination with Deep Learning
- DeepSphere: towards an equivariant graph-based spherical CNN
Cited by in corpus (9)
- Graph-Based Deep Learning for Medical Diagnosis and Analysis: Past, Present and Future
- Simplicial Neural Networks
- SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration
- Symmetric Spaces for Graph Embeddings: A Finsler-Riemannian Approach
- Group Equivariant Stand-Alone Self-Attention For Vision
- Physics-aware Spatiotemporal Modules with Auxiliary Tasks for Meta-Learning
- Deformation Robust Roto-Scale-Translation Equivariant CNNs
- Equivariant Networks for Pixelized Spheres
- ChebLieNet: Invariant Spectral Graph NNs Turned Equivariant by Riemannian Geometry on Lie Groups