The pytorch implemention of Cycle-gan

This is a Pytorch implementation of the "Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks" paper.

Dataset

you can download data from link , or run the script get_data.py to choose dataset you want download.

python get_data.py

facades: 400 images from the CMP Facades dataset. [Citation]
cityscapes: 2975 images from the Cityscapes training set. [Citation]
maps: 1096 training images scraped from Google Maps.
horse2zebra: 939 horse images and 1177 zebra images downloaded from ImageNet using keywords wild horse and zebra
apple2orange: 996 apple images and 1020 orange images downloaded from ImageNet using keywords apple and navel orange.
summer2winter_yosemite: 1273 summer Yosemite images and 854 winter Yosemite images were downloaded using Flickr API. See more details in our paper.
monet2photo, vangogh2photo, ukiyoe2photo, cezanne2photo: The art images were downloaded from Wikiart. The real photos are downloaded from Flickr using the combination of the tags landscape and landscapephotography. The training set size of each class is Monet:1074, Cezanne:584, Van Gogh:401, Ukiyo-e:1433, Photographs:6853.
iphone2dslr_flower: both classes of images were downlaoded from Flickr. The training set size of each class is iPhone:1813, DSLR:3316. See more details in our paper.

To train a model on your own datasets, you need to create a data folder as the following. (orange2apple is dataset name)

data/
	orange2apple/
			train/
				 A/...jpg
				 B/...jpg
			 test/
				 A/...jpg
				 B/...jpg

python cyclegan.py ARGS

Possible ARGS are:

--epochs number of epochs of training, default is 200;
--dataset_name name of the dataset, default is "orange2apple";
--batch_size size of the batches, default is 1;
--lr adam: learning rate, default is 0.0002;
--decay_epoch epoch from which to start lr decay, default is 100;
--img_height size of image height (default is 128);
--img_width size of image width (default is 128);
--sample_interval interval between saving generator outputs(default is 100) ;
--checkpoint_interval interval between saving model checkpoints (default is -1) ;
--n_residual_blocks number of residual blocks in generator (default is 9) ;

An example:

python cyclegan.py --n_residual_blocks 6