pGAN

This repository is unofficial implementation of pGAN with PyTorch.

A refined version of Original implementation.

Fix several problems

1. Code updated to newest version
2. Serveral bugs fixed
3. Better training and testing process

Training

python pGAN.py --dataroot datasets/IXI --name pGAN_run --direction BtoA --training

name - name of the experiment

direction - direction of synthesis. If it is set to 'AtoB' synthesis would be from data_x to data_y, and vice versa

Testing

python pGAN.py --dataroot datasets/IXI --name pGAN_run --direction BtoA --phase test --results_dir results/

name - name of the experiment

direction - direction of synthesis. If it is set to 'AtoB' synthesis would be from data_x to data_y, and vice versa

Citation

You are encouraged to modify/distribute this code. However, please acknowledge this code and cite the paper appropriately.

@article{dar2019image,
  title={Image Synthesis in Multi-Contrast MRI with Conditional Generative Adversarial Networks},
  author={Dar, Salman UH and Yurt, Mahmut and Karacan, Levent and Erdem, Aykut and Erdem, Erkut and {\c{C}}ukur, Tolga},
  journal={IEEE Transaction on Medical Imaging},
  year={2019},
  publisher={IEEE}
}

For any questions, comments and contributions, please contact Salman Dar (salman[at]ee.bilkent.edu.tr)

(c) ICON Lab 2019

Acknowledgments

This code is based on implementations by pGAN-cGAN and CycleGAN and pix2pix in PyTorch.