Abstract: We focus on learning descriptive geometry and motion features from 4D point cloud sequences in this work. Existing works usually develop generic 4D learning tools without leveraging the ...