Fast, Expressive SE(n) Equivariant Networks through Weight-Sharing in Position-Orientation Space Analysis

Weight-sharing in convolutional networks facilitates efficient processing of 3D point clouds with directional information.