Open sendeniz opened 3 years ago
The labels are loaded by ImageFolder in dataloader.py:
dataset = torchvision.datasets.ImageFolder(root=dataroot,transform=transform)
The ImageFolder Class can assign classification labels from folder names. Please check out the documentation
Dear otepencelik,
Thank you for sharing your work, its very interesting! I was curious about the cgan implementation. I have studied your code base and repo. In the cGAN, however I cannot find the corresponding line where you condition art generation on the art style labels. Where are art style labels loaded and passed to the model ? The dataloader.py loads images. It does not load labels. Then the cgan_model uses n_classes, which are the number of classes that exist in the data-set. During training you pass:
real_style_labels = real_style_labels.to(device)
where are they from ? Could you elaborate on this ? I really need to know.