Phillip Isola

From http://pytorch.org/docs/nn.html: > At evaluation time (.eval()), the default behaviour of the InstanceNorm module stays the same i.e. running mean/variance is NOT used for normalization. One can force using stored...

In my experience the important thing is to have the test setting exactly match the train setting. One way to do this is to use the same batch size at...

This is an interesting question. One thing to recognize first is that the _generator_ has receptive fields that cover the entire image, and that's what allows the generator to model...

For both 1 channel images and 4 channel images, you will need to edit `util.preprocess` and `util.deprocess` (the permutation assumes a 3 channel image). You will also need to edit...

Very strange! Can you share some examples of the input and output pairs you trained on? The discriminator does tend to be rather insensitive to minor translations, but this looks...

Code for calculating the receptive field is here, and see rest of thread for more info: https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/issues/39#issuecomment-336737526

You might not have permission to write to that directory. Can you try chmodding to give yourself permission?

Thanks for the suggestion, this is definitely on the to do list and we will try to provide it soon!

Thanks for finding this bug! I'll look into it when I get a chance. @sdlovecraft do you mind sharing your script in the meantime?

thanks!