Fetching the paper…

MViTv2: Improved Multiscale Vision Transformers for Classification and Detection · Around