Cross-view Action Modeling, Learning and Recognition

Jiang wang, Xiaohan Nie, Yin Xia +2

Existing methods on video-based action recognition are generally view-dependent, i.e., performing recognition from the same views seen in the training data. We present a novel mult…