falconbear/cartoon-gan-sample

Cartoon-GAN

Paper: https://arxiv.org/abs/2005.07702

Description

This project takes on the problem of transferring the style of cartoon images to real-life photographic images by implementing previous work done by CartoonGAN. We trained a Generative Adversial Network(GAN) on over 60 000 images from works by Hayao Miyazaki at Studio Ghibli.

To the people asking for the dataset, im sorry but as the material is copyright protected i cannot share the dataset.

Dependencies

1. Install Anaconda from https://www.anaconda.com/

2. Install pytorch at: https://pytorch.org/get-started/locally/

3. Install dependencies:

   python -m pip install tqdm pillow numpy matplotlib opencv-python
   

4. For predicting videos you will also need ffmpeg

Weights

Weights for the presented models can be found here

Training

All training code can be found in experiment.ipynb

Predict

Predict by running predict.py.

Example:

python predict.py -i C:/folder/input_image.png -o ./output_folder/output_image.png

Predictions can be made on images, videos or a folder of images/videos.

Demonstration

Image #	Original	CartoonGAN	GANILLA	Our implementation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

Citation

@misc{andersson2020generative,
      title={Generative Adversarial Networks for photo to Hayao Miyazaki style cartoons}, 
      author={Filip Andersson and Simon Arvidsson},
      year={2020},
      eprint={2005.07702},
      archivePrefix={arXiv},
      primaryClass={cs.GR}
}