Fetching the paper…

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