This is the repository for team "If it works, it works" and their submission to the CTP Cohort 9 hackathon.

The following people contributed to this project:

Yan Chen
Owen Liang
Fu Jun Pan
Ze Hong (Jason) Wu

The purpose of this project is to create a Generative Adverserial Network (GAN) that takes an input image and returns an (for lack of a better word) "animefied" output image, as well as an associated front end website to allow users to make use of the GAN without having to clone this repo and run things on the command line.

This project makes use of the following technologies:

Front-end: React.js, HTML/CSS
Back-end: Django (with REST framework), nginx, Python
Pre-trained GAN models: sourced from this repo
API endpoint: from this fork of the repo.
Paper that we based some of our work on: See References.

We attempted to train our own models using PyTorch, however the models generated did not live up to our expectations and we made the choice to use pre-trained models from another person's repo.

Follow the following steps to make use of our "animefier".

NOTE: The link displayed below no longer functions. This is because the app and server linked was hosted on the personal device of one of the team members and was taken offline after the hackathon judges inspected it and the hackathon concluded.

1. Visit this link.
2. Select an image that you want to "animefy" (using the "Browse" button) and select one of four "animefier" styles from the drop-down menu.
3. Click the "Upload" button and wait. Since the models in question are large and the back-end doesn't have a lot of resources, it might take some time to receive your output image.
4. The website will eventually return the output image. The image will not be permanently stored on our back-end and might be deleted at any moment. Because of this, we recommend that you save the image quickly by right-clicking and selecting "Save Image As".
5. Repeat Steps 1 through 4 for as many or few images as you want.

[1] Chen, Yang, Yu-Kun Lai, and Yong-Jin Liu. "CartoonGAN: Generative Adversarial Networks for Photo Cartoonization." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.

(Full paper: http://openaccess.thecvf.com/content_cvpr_2018/papers/Chen_CartoonGAN_Generative_Adversarial_CVPR_2018_paper.pdf)

[2] https://github.com/znxlwm/pytorch-CartoonGAN