Loading...
CycleGAN Implementation for Creating Modern Thai Houses
Author(s)
Kosavinta, Satakhun
Kunapinun, Alisa
Date Issued
January 1, 2023
Type
Conference Paper
Abstract
Nowadays, artificial intelligence (AI) plays a vital role in developing future trends and technology. Artificial intelligence approaches and applications have been developed over several decades to achieve AI systems for architectural creation to replace architectural design by humans at certain levels. However, the current development of AI models and architectural creation frameworks is inadequate and unsatisfactory. Moreover, limited studies on AI applications have focused on vernacular design or traditional architecture. To the best of our knowledge, this is the first study aiming to develop a deep learning technology that can design traditional architectures. This study aims to investigate the aspects of architectural design for vernacular architecture, while focusing on traditional Thai houses, and finding the proper types of generative adversarial networks (GANs) leveraged for architectural creation. By exploring its implications in the architectural field, this study proposes an approach of GANs for designing a specific architectural project style, and provides a model of CycleGAN in architectural creation using an empirical study on Thai vernacular houses as a case study. After training, the results show that the AI model can be applied to the design of modern Thai houses by enriching modern international houses with characteristics of traditional Thai houses. The primary transforming identities of the Thai vernacular architecture selected by the AI model exhibit a high gable (with top element), “Pan-lom” (windbreak on the gable), Thai-style platform (usually leaving the first floor as open space), and wood texture (including color and Thai-style assembly pattern).
Citation
Habits of the Anthropocene Scarcity and Abundance in A Post Material Economy Proceedings of the 43rd Annual Conference of the Association for Computer Aided Design in Architecture Acadia 2023, 2, 572-581, 2023