This repository contains an implementation of LocoGAN - Locally Convolutional GAN) in PyTorch, proposed by Łukasz Struski, Szymon Knop, Jacek Tabor, Wiktor Daniec, Przemysław Spurek (2020).
|-- src/ - contains an implementation of LocoGAN allowing to reproduce experiments from the original paper
|---- architectures/ - files containing architectures proposed in the paper
|---- factories/ - factories used to create objects proper objects base on command line arguments
|---- lightning_callbacks/ - directory contains PyTorch Lightning callbacks used during the experiments
|---- lightning_modules/ - directory contains PyTorch Lightning modules to conduct the experiments
|---- modules/ - custom neural network layers used in models
|---- tests/ - a bunch of unit tests
|---- train_locogan.py - the main script to run all of the experiments
|-- results/ - directory that will be created to store the results of conducted experiments
|-- data/ - default directory that will be used as a source of data and place to download datasets
Experiments are written in pytorch-lightning to decouple the science code from the engineering. For more details refer to PyTorch-Lightning documentation
The simplest experiment can be executed by running the following command in the src directory:
python -m train_locogan --dataroot <path_to_ffhq_thumbnails_dataset>
The code allows manipulating some of the parameters(for example using other versions of the model, changing learning rate values) for more info see the list of available arguments in src/args_parser.py file
To run the unit tests execute the following command:
python -m unittest
- python3
- pytorch
- torchvision
- numpy
- pytorch-lightning
FID Score can be computed with the following code and repository:
For the experiments presented in this repository we used the FFHQ dataset.
This implementation is licensed under the MIT License