video dialog 有影响力的论文详解

Posted 2020-12-14 jun0wanan

Multi-View Attention Networks for Visual Dialog

 

 *    切入点：问题的重点是哪个 ，需要充分理解才能抓住重点——给定问题的语义意图  determining the semantic intent of the given question and then leveraging question-relevant information from the dialog history and visual contents

这个卖点好，基于这个提出了三个module，提高了0.1%