c-he/Animoe

Anime Face Creation with Generative Adversarial Networks

Dataset

The original dataset can be downloaded from the Anime Face Dataset. We then parse the tags using Illustration2Vec and delete those failure cases.

Run the Code

To run the code, we need first run the file preprocess.py in the model folder:

python preprocess.py

It will dump all of the images and tags into one binary file, which can speedup our training process.

Then just run the code train.py with default parameters:

python train.py

The output images and checkpoint models are put in the folder output and we can visualize the loss with TensorBoard.

Web Application

Demo

LINK

Environment

Python virtual environment is recommended!

pip install Flask

Run

python3 webapp.py {your .pth model file path}

Example:

python3 webapp.py output/netG_epoch_200.pth

Web App Preview

open http://127.0.0.1:5000/

Notes

1. The anime eye icon in the logo is adapted from Wikipedia.