Video violence recognition and localization using a semi-supervised hard attention model

作者:

Highlights:

• Videos contain irrelevant information like background and surrounding objects.

• Isolating the valuable information in a video increases the classification accuracy.

• Reinforcement learning-based hard attention can identify critical visual information.

• Hard attention learns violence localization without explicit spatial annotation.

摘要

•Videos contain irrelevant information like background and surrounding objects.•Isolating the valuable information in a video increases the classification accuracy.•Reinforcement learning-based hard attention can identify critical visual information.•Hard attention learns violence localization without explicit spatial annotation.

论文关键词:Deep reinforcement learning,Violence detection,Hard attention,Video classification,Semi-supervised learning

论文评审过程:Received 19 April 2022, Revised 7 August 2022, Accepted 4 September 2022, Available online 7 September 2022, Version of Record 16 September 2022.

论文官网地址:https://doi.org/10.1016/j.eswa.2022.118791