Hello, I am interested to make one to many GAN images from one domain to another. I have added noise vector into input of UNET_256. Unfortunately, generated images vary but not much (added artifacts are lighter darker or vanished but usually the same). I have tried yours bicycleGAN, but it was made for pix2pix (paired) datasets. I tried to trick him with implemented unaligned DS but then after some epochs he started rather crossing images and it get messy. Do you have any idea which could help me to implement multimodal cycleGAN based on your repo ?
Hello, I am interested to make one to many GAN images from one domain to another. I have added noise vector into input of UNET_256. Unfortunately, generated images vary but not much (added artifacts are lighter darker or vanished but usually the same). I have tried yours bicycleGAN, but it was made for pix2pix (paired) datasets. I tried to trick him with implemented unaligned DS but then after some epochs he started rather crossing images and it get messy. Do you have any idea which could help me to implement multimodal cycleGAN based on your repo ?
Greetings