Johns Hohns
Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions (ICCV2021)