Towards Generating Image Assets Through Deep Learning for Game Development
Adam Richard Tilson, Craig M. Gelowitz · 2019
This paper outlines some preliminary research toward the viability of utilizing unsupervised generative deep learning networks, such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), to create image assets for game development. This work explores existing GAN research and the viability of using generative networks to generate both textures and human face assets within a game engine. This research and its preliminary results examines the viability and advantages/disadvantages of utilizing generative networks directly within game engines. It also examines the potential trade-offs of a generative approach versus more traditional approaches for image assets in video game development.