I am working with images 400x600, and trained a resnet6 pix2pix with --preprocess None. The output images are 200x600, and it is the left cropped image from the former 400x600 image, even though I specify --preprocess None during testing. Any help?
In the pix2pix setting, we assume that the input data contains both image_A and image_B. The program will split the input into two (see this line). In your case, your 400x600 image is split into 200x600 as input and 200x600 as output. We provided a script to prepare the paired dataset. (see here)
I am working with images 400x600, and trained a resnet6 pix2pix with --preprocess None. The output images are 200x600, and it is the left cropped image from the former 400x600 image, even though I specify --preprocess None during testing. Any help?