I tried implementing CycleGAN for image to image conversion of photo images to monet style images for a kaggle competition.
But everytime it is generating the same images and those images are all gray.
I am not getting what mistake have i done.
Please guide me how to correct this.
This is a link to my notebook cyclegan_notebook
hey @namayevivek87 Im sorry that I dont have an answer to your question. But I have a doubt, I have been trying to implement cyclegan for a while. What I don’t understand is why we have to use a specific class called “cyclegan_dataset” to load the data. Is it not possible to use pytorch functions like Imagefolder or Dataloader to load the data?
The reason why your output images are gray is because you have only used one discriminator and one generator for both monet and real. You need to have two discriminators and two generators.
You said that u were also trying to implement cyclegans right?
So did u complete it ? Or did u find any useful resource ?
If so can please send the notebook link, I am also looking for help for this but most of sources I found online were using tensorflow .