Multi-Modal fusion with multi-level attention for Visual Dialog

作者：

Highlights：

• We propose a novel visual dialog method with multi-level attention.

• Three high-level attention modules are devised to select important words.

• We also use attention to select relevant regions in the image.

• We show the multi-level attention is effective in the visual dialog.

摘要

•We propose a novel visual dialog method with multi-level attention.•Three high-level attention modules are devised to select important words.•We also use attention to select relevant regions in the image.•We show the multi-level attention is effective in the visual dialog.

论文关键词：Visual Dialog,Multi-Modal,Multi-Level,Attention mechanism,00-01,99-00

论文评审过程：Received 10 July 2019, Revised 7 September 2019, Accepted 24 October 2019, Available online 11 November 2019, Version of Record 6 May 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2019.102152