Hi, great to see the interesting work!
I'm interested in the synthesized images generated from a random noise
have you tried to training any matting models using purely images generated from GAN?
Thanks for your interest, we use the StyleGAN pretrained on the face dataset, it can only synthesize faces but not the face-matte pairs, so we cannot train the matting models using purely generated images.
Hi, great to see the interesting work! I'm interested in the synthesized images generated from a random noise have you tried to training any matting models using purely images generated from GAN?