Publications / 2024 Proceedings of the 41st ISARC, Lille, France

Training and Generating Apartment Plan Graphic Images for Commercial Uses

Taesik Nam, Youngchae Kim, Sumin Chae, Youngjin Yoo, Jin Kook Lee
Pages 1299-1302 (2024 Proceedings of the 41st ISARC, Lille, France, ISBN 978-0-6458322-1-1, ISSN 2413-5844)
Abstract:

This paper introduces a methodology utilizing image generation AI to create 3D floor plan images for Korean-style apartments. The growing interest in spatial dynamics, driven by changes in residential preferences, has elevated the importance of widespread comprehension and the demand for floor plans. Specifically, apartment floor plans, crucial for commercial purposes, require visually compelling representations that are both diverse and detailed yet easily comprehensible. Traditional 2D floor plans face limitations in public understanding due to the use of architectural terminology and specialized symbols. As an alternative, 3D floor plan images offer a comprehensive understanding of three-dimensional structures and arrangements, with potential applications across various commercial sectors. The recent emergence of image generation AI presents opportunities for the automated creation and alteration of architectural images. Fine-tuning the generated results can be achieved through detailed prompt settings and additional training techniques. The methodology for constructing a 3D floor plan image generation model for Korean-style apartments involves: 1. Creating a 3D Building Information Modeling (BIM) model based on Korean-style apartment floor plans and extracting 3D floor plan images based on different areas. 2. Training the floor plan image generation AI model by combining the extracted images with text and utilizing the trained model to automatically generate 3D floor plan images. The resulting images can find applications not only in architecture but also in real estate, design, advertising, and other diverse commercial fields. This approach suggests the potential development of a versatile model applicable across various domains, especially when integrated with generative models.

Keywords: Architectural Visualization, 3D Floor Plan, Image Generation AI