This repository is a simplified CycleGAN implementation aiming for image translation such as horse-to-zebra(may has some bugs).
The official implementation is here
.
├── dataIO // the code of dataset I/O module
│ ├── __init__.py
│ ├── datasets.py
│ ├── preprocess.py
│ └── utils.py
├── model // the code of definition of CycleGAN model
│ ├── CycleGAN.py
│ ├── __init__.py
│ ├── basic_blocks.py
│ ├── networks.py
│ └── utils.py
├── weights // the pretrained weight files
├── test.py // the code for testing
└── train.py // the code for training
use python train.py --dataroot ./data/horse2zebra [OPTIONS]
some of the options are listed below, you can use python train.py -h for more information.
--dataroot the root path of the dataset, must have 4 subfolders: trainA, trainB, testA, testB
--weight_file where to store the temporary weights
--res_layers the num of Resnet blocks in CycleGan
--batch_size the batch size of input images
--shuffle whether to shuffle input images
--crop_size the crop size
--crop_pos the crop position
--flip whether to flip input images horizontally
--grayscale whether to read images as gray images
--resize the size of images after resizing
--lr the learning rate
--gpu whether to use gpu to train
--epochs the num of training epochs
use python test.py --dataroot ./your/dataset [OPTIONS] and the result will reside in dataroot/fake/.
most options are same as training options except --lr --epochs, use python test.py -h for more information.