seominlee / creative-adversarial-networks

(WIP) Implementation of Creative Adversarial Networks https://arxiv.org/pdf/1706.07068.pdf

Creative Adversarial Networks

128x128 pixel Samples from CAN train on WikiART.

A WIP implementation of CAN: Creative Adversarial Networks, Generating "Art" by Learning About Styles and Deviating from Style Norms. Repo bases DCGAN implementation on DCGAN-tensorflow with modifications to reduce checkerboard artifacts according to this distill article

The paper authors basically modified the GAN objective to encourage the network to deviate away from art norms.

Getting the Dataset

We used this compiled wikiart dataset available here. Using the dataset is subject to wikiart's terms of use

Extract the dataset, then set the path in train.sh

Training a DCGAN model

Edit the parameters of train.sh then

bash train.sh

Citation

If you use this implementation in your own work please cite the following

@misc{2017cans,
  author = {Phillip Kravtsov and Phillip Kuznetsov},
  title = {Creative Adversarial Networks},
  year = {2017},
  howpublished = {\url{https://github.com/mlberkeley/Creative-Adversarial-Networks}},
  note = {commit xxxxxxx}
}

Authors

Phillip Kravtsov

Phillip Kuznetsov