Transformer feature collapse of Temporal Action Detection via Multi-granularity Semantic Enhancement

Open Webpage

Xin An, Peng Zhao, Guiqin Wang, Cong Zhao, Shusen Yang

Published: 2025, Last Modified: 04 Aug 2025Neurocomputing 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Mitigate temporal feature collapse issue for TAD.•Model long temporal dependency for multi-instance videos.•Enhance feature representations from diverse semantic spaces.•Expensive experiments prove the effectiveness of our method.