Multi-stream Global-Local Motion Fusion Network for skeleton-based action recognition

Yanpeng Qi, Chen Pang, Yiliang Liu, Lei Lyu

Abstract: Highlights•A multi-stream model Global–Local Motion Fusion Network is proposed.•Grouping GCN aims to enforce the ability to aggregate local spatial information.•Spatial Self-attention aims to extract spatial long-term motion relationships.•Temporal Self-attention aims to capture temporal long-term motion relationships.