Identifying the key frames: An attention-aware sampling method for action recognition

作者：

Highlights：

• We propose an attention-aware sampling method to select discriminative frames in videos, where the agent is trained by deep reinforcement learning.

• Identifying key frames can be taken as a weakly supervised problem, therefore, we also generate pseudo labels to train the agent together with the reward supervision.

• We conduct experiments on two widely used benchmark datasets to demonstrate the effectiveness of our method and achieve competitive results.

摘要

•We propose an attention-aware sampling method to select discriminative frames in videos, where the agent is trained by deep reinforcement learning.•Identifying key frames can be taken as a weakly supervised problem, therefore, we also generate pseudo labels to train the agent together with the reward supervision.•We conduct experiments on two widely used benchmark datasets to demonstrate the effectiveness of our method and achieve competitive results.

论文关键词：Action recognition,Deep learning,Reinforcement learning,Pseudo labels

论文评审过程：Received 22 December 2020, Revised 24 April 2022, Accepted 17 May 2022, Available online 19 May 2022, Version of Record 1 June 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108797