Generating Gameplay-Relevant Art Assets with Transfer Learning
This addresses a domain-specific need for game developers to reduce costs in creating gameplay-relevant art assets, but it is incremental as it builds on existing image generation methods.
The paper tackled the problem of generating game visuals that incorporate gameplay mechanics, using a Convolutional Variational Autoencoder system with Pokémon sprites and type information, resulting in improved visual quality and stability over unseen data through transfer learning.
In game development, designing compelling visual assets that convey gameplay-relevant features requires time and experience. Recent image generation methods that create high-quality content could reduce development costs, but these approaches do not consider game mechanics. We propose a Convolutional Variational Autoencoder (CVAE) system to modify and generate new game visuals based on their gameplay relevance. We test this approach with Pokémon sprites and Pokémon type information, since types are one of the game's core mechanics and they directly impact the game's visuals. Our experimental results indicate that adopting a transfer learning approach can help to improve visual quality and stability over unseen data.