A clean and readable Pytorch implementation of CycleGAN (https://arxiv.org/abs/1703.10593)
Code is intended to work with Python 3.6.x
, it hasn't been tested with previous versions
To plot loss graphs and draw images in a nice web browser view
pip3 install visdom
Before training:
python -m visdom.server
First, you will need to download and setup a dataset. The easiest way is to use one of the already existing datasets on UC Berkeley's repository:
./download_dataset <dataset_name>
Valid <dataset_name> are: apple2orange, summer2winter_yosemite, horse2zebra, monet2photo, cezanne2photo, ukiyoe2photo, vangogh2photo, maps, cityscapes, facades, iphone2dslr_flower, ae_photos
Dataset can be downloaded from here.
Alternatively you can build your own dataset by setting up the following directory structure:
.
├── datasets
| ├── <dataset_name> # i.e. brucewayne2batman
| | ├── train # Training
| | | ├── A # Contains domain A images (i.e. Bruce Wayne)
| | | └── B # Contains domain B images (i.e. Batman)
| | └── test # Testing
| | | ├── A # Contains domain A images (i.e. Bruce Wayne)
| | | └── B # Contains domain B images (i.e. Batman)
./train --dataroot datasets/<dataset_name>/ --cuda
This command will start a training session using the images under the dataroot/train directory with the hyperparameters that showed best results according to CycleGAN authors. You are free to change those hyperparameters, see ./train --help
for a description of those.
Both generators and discriminators weights will be saved under the output directory.
You can also view the training progress as well as live output images by running python3 -m visdom
in another terminal and opening http://localhost:8097/ in your favourite web browser. This should generate training loss progress as shown below (default params, horse2zebra dataset):
./test --dataroot datasets/<dataset_name>/ --cuda
This command will take the images under the dataroot/test directory, run them through the generators and save the output under the output/A and output/B directories. As with train, some parameters like the weights to load, can be tweaked, see ./test --help
for more information.
Examples of the generated outputs (default params, horse2zebra dataset):