Pretrained model of BebyGAN for upsacling factor 2 and 3

[Techzist16]

Please provide us the pretrained model of BebyGAN for upsacling factors 2 and 3.

[fenglinglwb]

Sorry. Following previous work, we only conducted experiments on the x4 setting.

[Techzist16]

@fenglinglwb Is it possible to train the model for upscaling factors 2 and 3 on our own?

[fenglinglwb]

Sure. You could first train an RRDB backbone using L1 loss for initialization, and then train the initialized model using GAN loss.

[Techzist16]

@fenglinglwb Thank you so much for your prompt reply. What changes do we need to make to the training code if we want to train the model on our own dataset?