Generating video game characters using StyleGAN2

Abstract

GANs have been getting better and better each year. The state of the art GAN models for generating 2D images have become so good it is hard to differentiate generated images nowadays. In this paper we create 3 different sparse data sets from video game assets and train them with StyleGAN2 to generate new artwork based on the previously existing artworks of the video game in question

Keywords

• Generative art
• Videogame
• StyleGAN

References

1. BRADISKI, G. (2000). The OpenCV Library. Dr. Dobb’s Journal of Software Tools, 2000.
2. Karras et al. Analyzing and improving the image quality of StyleGAN.
3. Karras et al. A style-based generator architecture for generative adversarial networks.
4. Karras et al. Training generative adversarial networks with limited data.
5. The ImageMagick Development Team. Imagemagick.