Architectural Styles through the Lens of Artificial Intelligence: A Review

Abstract

Architectural style is a topic with a long history. Different styles reflect the diversity of architecture, carry the imprint of the times and regions, and bear important aesthetic and cultural value. With the advent of artificial intelligence (AI), new methodologies for classifying these styles have emerged, offering a fresh perspective on architectural analysis. This article provides a review of the AI-based architectural style classification research, examining research topics, datasets, data types, and algorithms employed in the field. The core of the current architectural style classification algorithm is supervised learning, with convolutional neutral network most frequently used. Most research use images or graphs to represent architecture information. For future research, a larger standardized dataset is needed, and unsupervised learning should be given more attention. AI-based architectural style classification may be used in areas such as personalized architectural recommendations and new style generation, bringing a new perspective to architectural design.