WSI-cycleGAN

Based on the Tensorflow implementation for learning an image-to-image translation without input-output pairs. The method is proposed by Jun-Yan Zhu in Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks.

This code was forked from https://github.com/xhujoy/CycleGAN-tensorflow and modified to work optimally on Whole Slide images WSIs by Brendon Lutnick

You can download the pretrained model from this url and extract the rar file to ./checkpoint/.

Prerequisites

tensorflow r1.1
numpy 1.11.0
scipy 0.17.0
pillow 3.3.0

Getting Started

Train

Download a dataset (e.g. zebra and horse images from ImageNet):

bash ./download_dataset.sh horse2zebra

Train a model:

CUDA_VISIBLE_DEVICES=0 python main.py --dataset_dir=horse2zebra

Use tensorboard to visualize the training details:

tensorboard --logdir=./logs

Test

Finally, test the model:

CUDA_VISIBLE_DEVICES=0 python main.py --dataset_dir=horse2zebra --phase=test --which_direction=AtoB

Training and Test Details

To train a model,

CUDA_VISIBLE_DEVICES=0 python main.py --dataset_dir=/path/to/data/

Models are saved to ./checkpoints/ (can be changed by passing --checkpoint_dir=your_dir).

To test the model,

CUDA_VISIBLE_DEVICES=0 python main.py --dataset_dir=/path/to/data/ --phase=test --which_direction=AtoB/BtoA

Datasets

This code takes 2 folders containing WSIs as an input. The code will pull random patches from the slides at runtime and attempt to translate bnetween the two sets. For example you could place H&E WSIs in folder A and PAS WSIs in folder B.

model.py init contains 2 variables

self.dataset_dirA

and

self.dataset_dirB

which should be changed.

The custom Tensorflow input pipeline used for this network can be downloaded here

Reference

The torch implementation of CycleGAN, https://github.com/junyanz/CycleGAN
The tensorflow implementation of pix2pix, https://github.com/yenchenlin/pix2pix-tensorflow

SarderLab/WSI-cycleGAN