PCMG:3D point cloud human motion generation based on self-attention and transformer

Weizhao Ma, Mengxiao Yin, Guiqing Li, Feng Yang, Kan Chang

Abstract: Previous methods for human motion generation have predominantly relied on skeleton representations to depict human poses and motion. These methods typically use a series of skeletons to represent the motion of a human. However, they are not directly suitable for handling the 3D point cloud sequences obtained from optical motion capture. To address this limitation, we propose a novel network called point cloud motion generation (PCMG) that can handle both skeleton-based motion representation and point cloud data from the human surface. PCMG is trained on finite point cloud sequences and is capable of generating infinite new point cloud sequences. By providing a predefined action label and shape label as input, PCMG generates a point cloud sequence that captures the semantics associated with these labels. PCMG achieves comparable results to state-of-the-art methods for action-conditional human motion generation, while outperforming previous approaches in terms of generation efficiency. The code for PCMG will be available at https://github.com/gxucg/PCMG