Neural style transfer with CycleGANs

For our final project, we make use of cyclic generative adversarial networks (GANs) to produce new images from source images.
Explore the docs »

Report Bug · Request Feature

Table of Contents

1. About The Project
2. Getting Started
   • Prerequisites
   • Usage
   • Repository Structure
3. Contributing

About The Project

In our 3-person team, we developed a Cycle GAN that can convert human to anime faces and vice versa (with some success).

It was a group project as part of the graduate course DD2424 at the Royal Institute of Technology, Stockholm. The overarching purpose of the project was to get a better understanding of deep learning technologies.

For a written version of our results, see the attached article.

(back to top)

Getting Started

Prerequisites

Install the

requirements.txt

Usage

We ran all of our experiments on Google Cloud GPUs. If you wish to do so, too, find the setup descriptions in

setting_up_vm.txt

(back to top)

Repository structure

• data: the respective images for the cycle GAN
• models / saved_images / stats: will be populated on Google Cloud VM
• src:
  • device: see, if cuda is available
  • discriminator / generator: our Cycle GAN architecture
  • load_data: the data loader
  • train: the training script and training parameters
  • utils: checkpointing and other utility functions

(back to top)

Contributing

If you have a suggestion that would make this better, please fork the repo and create a pull request. You can also simply open an issue with the tag "enhancement". Don't forget to give the project a star! Thanks again!

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

(back to top)