Closed diamond0910 closed 1 year ago
I am not sure I fully understand the question. When we encode the segmentation map into StyleGAN's latent space, we are in a sense mapping the segmentation map to an encoding of a real face image. The output of the task is a real image, not a segmentation map.
Thank you for your great work.
In the paper, you show you can encode the mask to the first seven w+ latant, and you random latent for the rest layers. But I see your code. The code directly encodes the mask to the whole w+ latent. Why? This will lead that the style-mixing operation generates different masks.
Thank you.