BriefGPT.xyz
Jul, 2021
局部到全局的视觉Transformer自注意力
Local-to-Global Self-Attention in Vision Transformers
HTML
PDF
Jinpeng Li, Yichao Yan, Shengcai Liao, Xiaokang Yang, Ling Shao
TL;DR
本研究提出了多通路结构的Transformer模型，实现局部到全局的多粒度特征推理，相较于现有的分层设计模型，在增加了极小的计算量的同时，在图像分类和语义分割任务上取得了显著的提高。
Abstract
transformers
have demonstrated great potential in
computer vision
tasks. To avoid dense computations of
self-attentions
in high-resolution
→