Fine-gained Motion Enhancement for action recognition: Focusing on action-related regions

Xiong Gao, Zhaobin Chang, Yande Li, Xingcheng Ran, Wei Ke, Yonggang Lu

Abstract: Highlights•We propose an IDME module to focus more on the foreground action-related regions.•We propose a MMST module to accurately localize moving objects.•The MMST and IDME modules are simple, lightweight, and plug-and-play.•By combining the IDME with MMST, the FMENet achieves excellent performance.