Hello, when I use swin transformer as an image processor, the image input during the test can only be the same size as the image input for training, and the mask part needs to be changed. Will network performance be affected if I change the mask part code to a multiple of 8
Hello, when I use swin transformer as an image processor, the image input during the test can only be the same size as the image input for training, and the mask part needs to be changed. Will network performance be affected if I change the mask part code to a multiple of 8