no code implementations • 19 Jan 2021 • Chang Li, Qian Huang, Xing Li, Qianhan Wu
We employ depth motion images (DMI) as the templates to generate the multi-scale static representation of actions.
Action Classification Action Recognition +1