It was a good paper, and thanks for releasing the code. I have a question in terms of loss computation for discriminators. In the implementation, you have two discriminator losses, D_A and D_P, but I do not know what these two losses represent. Do D_A and D_P represent a loss for the global and the local discriminator, respectively? Additionally, in our project, we would like to have quantitative results by disabling the local discriminator. To do that, are we able to just set patchD_3 to 0? Or do we need to set patchD_3 and patchD to 0 and False?
Hi,
It was a good paper, and thanks for releasing the code. I have a question in terms of loss computation for discriminators. In the implementation, you have two discriminator losses, D_A and D_P, but I do not know what these two losses represent. Do D_A and D_P represent a loss for the global and the local discriminator, respectively? Additionally, in our project, we would like to have quantitative results by disabling the local discriminator. To do that, are we able to just set patchD_3 to 0? Or do we need to set patchD_3 and patchD to 0 and False?
Thanks,