Fetching the paper…

CLIPSelf: Vision Transformer Distills Itself for Open-Vocabulary Dense Prediction · Around