Técnicas de inteligência artificial para geração de insumos 2d

Abstract

The current artistic landscape in the production of video-games and animation went through radical change with the computer graphics revolution before the turn of the millennium. The ubiquitous presence of hand drawn animation in the theaters and games made with, also hand drawn, sprites gave place to production that made extensive, and sometimes exclusive, of 3D computer generated imagery and 2D games that now find themselves limited to a small niche due to the elevated costs and reduced consistency in 2D resources, since the tools for making them didn't evolve nearly as much as their 3D counterparts. This work then presents the development and implementation of two techniques aimed to help artists and designers in the development of 2D resources, specifically through the use of Generative Adversarial Networks for the generation of sprites and through the use of Genetic Algorithms to help automate level design. Despite the moderate success obtained in the sprite generation, due to the limitations of computational power, the implementation of genetic algorithm proved itself to be extremely efficient in helping the level design process. Through the work done it becomes possible to estimate that the constant advances in the availability of computational resources will make the use of Generative models as well as Genetic algorithms for 2D asset creation a reliable tool in the 2D designer and artist's toolkit.