Twins: Revisiting the Design of Spatial Attention in Vision Transformers

Twins：重新思考视觉Transformer中的空间注意力设计.

All Tokens Matter: Token Labeling for Training Better Vision Transformers

LV-ViT：使用标志标签更好地训练视觉Transformers.

Incorporating Convolution Designs into Visual Transformers

CeiT：将卷积设计整合到视觉Transformers中.

CvT: Introducing Convolutions to Vision Transformers

CvT：向视觉Transformer中引入卷积.

Per-Pixel Classification is Not All You Need for Semantic Segmentation

MaskFormer：逐像素分类并不是语义分割所必需的.

Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions

PVT：一种无卷积密集预测的通用骨干.