Video Human Action Recognition by the Fusion of Significant Motion and Coarse-fine Temporal Granularity Features
Kaishi Xu1*, Hanhua Cao2, Yang YI2