Portal de Periódicos da CAPES

Synthesizing Pokémon Trading Cards Using Nvidia StyleGAN2 from Home

2022; Springer International Publishing; Linguagem: Inglês

10.1007/978-3-031-16078-3_44

2367-3370

Kalvin Scoon, Kamil Samara,

Artificial Intelligence in Games

Artificial intelligence plays a large role in the way that we consume content acting as a director behind the scenes controlling what and how we interact with digital mediums daily. Only within the last decade has AI software and hardware advanced to a point where it can also help contribute directly to content creation with the use of technologies such as generative adversarial networks (GANs). In this paper, we will examine some modern GAN solutions as well as the viability of an at-home model training experience in 2021 and into 2022. Specifically, the task being examined is the image synthesis of Pokémon Trading Cards. This dataset was chosen for its familiarity, large number of uniform samples, accessibility, and to examine the effects of training on datasets that share many consistent elements (card layout, formatting, color schemes). Some implications of being able to generate images in such a way are faster prototyping, a more streamlined workflow, or near-instantaneous content creation for creative professionals and illustrators.