Multi-person Point Cloud Detection Transformer for Multi-Person Motion Classification

Simple DL project for 23短-课程综合实践II-深度学习前沿技术与应用 with T.Muller and Starg.

• We propose a novel framework for multi-person action classification in point cloud videos. To our knowledge, this is the first work to consider the multi-person action classification in point cloud videos.
• We adopt a sublayer of our network to split the point cloud videos into two parts, each of which represents one of the two participants. This is important for multi-person action classification.
• Our proposed model outperforms the state-of-the-art methods on SBU Kinect Interaction Dataset.