phillipi / pix2pix

Image-to-image translation with conditional adversarial nets

Home Page:https://phillipi.github.io/pix2pix/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

About the size of output image of generator

schrodingercatss opened this issue · comments

I want to train pix2pix coloring model with 256 * 256 pictures, and then I can color pictures of any size. How can I make the generator output pictures of any size?

It depends on the generator architecture. The UNet generator used in the original pix2pix only supports images with height and width divisible by 256. You can try other kinds of architecture. For example, the Resnet-based generator supports images with height and width divisible by 4.