PRA-Net: Point Relation-Aware Network for 3D Point Cloud Analysis

Learning intra-region contexts and inter-region relations are two effective strategies to strengthen feature representations for point cloud analysis. However, unifying the representation is not fully emphasized in existing methods. To this end, we propose a novel framework named Point Relation-Aware Network (PRA-Net), which composed of an Intra-region Structure (ISL) module Inter-region Relation (IRL) module. The ISL can dynamically integrate local structural information into features, while IRL captures adaptively efficiently via differentiable region partition scheme representative point-based strategy. Extensive experiments on several 3D benchmarks covering shape classification, keypoint estimation, part segmentation have verified effectiveness generalization ability PRA-Net. Code will be available at https://github.com/XiwuChen/PRA-Net .