Stacked Attention Networks for Image Question Answering

Zichao Yang, Xiaodong He, Jianfeng Gao +2

This paper presents stacked attention networks (SANs) that learn to answer natural language questions from images. SANs use semantic representation of a question as query to search…