README.md

RGAN-based-Pix2pix

This is the code of my conference paper "Image-to-Image Translation using a Relativistic Generative Adversarial Network " in the 11th International Conference of Digital Image Processing (ICDIP 2019). This code refers to the basic pix2pix-pytorch code heavily and thinks for the sharing.

We use a relativistic generative adversarial network to implement image domain transformation tasks, for example label-to-photo and photo-to-map.

Label-to-photo task on the Cityscapes dataset

Photo-to-map task on the Google-maps

Run

Requirements:

• Install PyTorch (tested on release 0.4.0 and 0.4.1).
• Clone Pix2pix-Pytorch as backbone training framework.

Training and Validation:

• Copy pix2pix_model.py, into pytorch-CycleGAN-and-pix2pix/models/ to exchange the basic GAN model.

• Launch training with Pix2pix-Pytorch as backbone training framework.