3DInAction: Understanding Human Actions in 3D Point Clouds

Yizhak Ben-Shabat; Oren Shrout; Stephen Gould

doi:10.1109/cvpr52733.2024.01888

CVPR2024顶会

3DInAction: Understanding Human Actions in 3D Point Clouds

Yizhak Ben-Shabat, Oren Shrout, Stephen Gould

DOI 出版方

2024年份

6顶会引用

摘要

3dincaction Figure 1. t-patches for action recognition. We propose a new representation for dynamic 3D point clouds. Termed t-patches, these are locally evolving point cloud sets aggregated over time. Learning features over t-patches provides an improved temporal point cloud representation for action understanding.