1 paper
Zhuoxuan Peng, Yiyi Ding, Yang Lin +1
Many modern video-based human action recognition (HAR) approaches use 2D skeleton as the intermediate representation in their prediction pipelines. Despite overall encouraging resu…