To address this, we propose a pyramid convolutional vision transformer (PCViT) that gets rid of the limitations of existing transformer methods. First, we employ a pyramid architecture to effectively ...