Gans Used to Generate 2d Game Resources

School of Computing and Data Science, Xiamen University Malaysia, Sepang, Malaysia

This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

Game Assets Generation (GAG) is to create all kinds of resources a game need, including game pictures and level design. Here we focus on the game pictures. With the development of machine learning, the generation of game assets may not be created by artists by Unity Technologies, Unreal Engines or other tools. Different types of Generative Adversarial Network (GANs) can be used generate 2D game resources. In the article, based on wide references, GANs are separated into 2 basic categories, which are direct generation GAN and Indirect generation GAN. And for game assets generation, there are different missions. The basic generation of game assets, texture generation, different angles generation and style transfer. Based on the logic, 4 typical GANs of precedents are introduced and compared in algorithms and effects in detail. Since different GANs are suitable for different missions, the most suitable public datasets are chosen to evaluate the performance of each method. Additionally, problems of the algorithms are discussed, following with some future research directions.