Past is important: Improved image captioning by looking back in time

作者：

Highlights：

• In the paper, we first propose a Visual Reserved model that enables previous visual context to be considered for the current sequence reasoning.

• Next, a Attentional-Fluctuation Supervised model is also proposed in reinforcement learning structure.

摘要

•In the paper, we first propose a Visual Reserved model that enables previous visual context to be considered for the current sequence reasoning.•Next, a Attentional-Fluctuation Supervised model is also proposed in reinforcement learning structure.

论文关键词：Image captioning,Reinforcement learning,Visual attention

论文评审过程：Received 27 January 2020, Revised 29 December 2020, Accepted 24 January 2021, Available online 10 February 2021, Version of Record 26 February 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116183