HFA-GTNet: Hierarchical Fusion Adaptive Graph Transformer network for dance action recognition

Ru Jia, Li Zhao, Rui Yang, Honghong Yang, Xiaojun Wu, Yumei Zhang, Peng Li, Yuping Su

Abstract: Highlights•A novel attention mechanism is designed to model spatial correlations.•A three-branch module is proposed to capture temporal relations among all frames.•An component is introduced to handle the spatial and temporal variations.•Two different types dance action datasets are constructed.