BriefGPT.xyz
Jul, 2023
关键词感知的视频问答的相对时空图网络
Keyword-Aware Relative Spatio-Temporal Graph Networks for Video Question Answering
HTML
PDF
Yi Cheng, Hehe Fan, Dongyun Lin, Ying Sun, Mohan Kankanhalli...
TL;DR
该论文提出了一种关键词感知的相对时空图网络(KRST)用于视频问答,通过在问题编码过程中使用注意机制让问题特征对关键词敏感,指导视频图构建,并整合了相对关系建模以更好地捕捉物体节点之间的时空动态,实验证明KRST方法在多个现有方法上具有优势。
Abstract
The main challenge in
video question answering
(VideoQA) is to capture and understand the complex spatial and temporal relations between objects based on given questions. Existing
graph-based methods
for VideoQA
→