ArtGan
General Idea: Translating real images to Realism art style images. This art style dates back to the 17th century and there is no image-to-image pair data available to train traditional networks, So here I am implementing ideas expressed in the paper Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks (https://arxiv.org/pdf/1703.10593v7.pdf) and creating a GAN to transfer real Image to realism art style image, a bi-product of this implementation will also be capable to convert Realism art style images to somewhat real photo.
I’m using Realism image folder from dataset https://github.com/cs-chan/ArtGAN/tree/master/WikiArt%20Dataset
All the datset used is probably copyrighted, please check Copyright & privacy policy of WikiArt for more information (https://www.wikiart.org/en/terms-of-use).
Only can be used for academic and research.
Sample from dataset:
Proposed Evaluation
Because of the above-mentioned issues, I’ll be deviating from baseline and evaluation metrics from the base paper and instead will be using a likeness score as defined in the paper A Novel Measure to Evaluate Generative Adversarial Networks Based on Direct Analysis of Generated Images (https://arxiv.org/pdf/2002.12345.pdf) which defines a likeness score as a combination of Creativity, Inheritance, and Diversity.