← Do Vision Transformers See Like Convolutional Neural Networks Awesome weakly supervised semantic segmentation→