학술논문
Visual Recognition for ZELDA Content Generation via Generative Adversarial Network
Document Type
Conference
Author
Source
2023 3rd International Conference on Artificial Intelligence (ICAI) Artificial Intelligence (ICAI), 2023 3rd International Conference on. :76-81 Feb, 2023
Subject
Language
Abstract
In video games, procedural content generation has a strong history. Current procedural content generation strategies, such as search-based, solver-based, rule-based, and language-based techniques, have been used to create levels, maps, character models, and surfaces in games. There has been a research area dedicated to game content generation. More recently, Generative tasks have been in charge of a wide range of content creations that are relevant to games. Although some front-line Generative Adversarial Networks (GANs) are used independently, others are used in conjunction with more traditional techniques or an intel-ligent environment. GANs model suffers from a problem known as mode collapse where duplicate content is generated. In this ar-ticle, we have applied a simple Generative Adversarial Network, Deep convolutional Generative Adversarial Network(DCGAN), and Wasserstein Generative Adversarial Network(WGAN) to the ZELDA data set for levels of content generation and conclude the results of the basis of visual recognition. Results show that WGAN generates visually good content.