BriefGPT.xyz
Jul, 2022
自监督视频表示学习的静态和动态概念
Static and Dynamic Concepts for Self-supervised Video Representation Learning
HTML
PDF
Rui Qian, Shuangrui Ding, Xian Liu, Dahua Lin
TL;DR
该研究提出了一种新的自监督视频表示学习方案，分别学习全局视觉概念和局部特征，使用交叉注意力机制聚合不同概念的详细本地特征来执行局部概念对比，并取得了UCF-101、HMDB-51和Diving-48的最新成果。
Abstract
In this paper, we propose a novel learning scheme for self-supervised
video representation
learning. Motivated by how humans understand videos, we propose to first learn general
visual concepts
then attend to dis
→