단독 주택 외관 시각화를 위한 생성형 인공지능기반 건축가 스타일 추가학습 모델 구축 및 활용 방안 = Generative AI-Based Construction of Architect’s Style-trained Models and its Application for Visualization of Residential Houses

다국어 초록 (Multilingual Abstract)

In the design process, visualization materials, including perspectives, play a crucial role in facilitating communication among stakeholders. However, their production requires a significant investment of time and resources. This paper suggests an architectural visualization method based on Generative AI for the initial design phase, incorporating a fine-tuning approach to streamline the visualization process. Over 10,000 residential house images reflecting architects’ styles were generated to assess the existing model‘s performance. Architects with lower recognition rates underwent training after constructing corresponding datasets. Subsequently, we demonstrated the practical application of the fine-tuned model through three scenarios. As a result, it was confirmed that fine-tuning effectively enhances the model’s performance. Moreover, the proposed methodology allows users to build personalized models based on individual preferences and generate architectural visualizations within seconds using prompts or simple seed images. This study would serve as a foundational contribution to further research in Generative AI-based early design visualization technology.