이미지 생성 AI 파인튜닝과 활용 사례 - 실내 공간 상세 스타일 키워드를 중심으로 - = Image Gen AI-based Model Fine-Tuning and Design Application- Focusing on Detailed Interior Design Style Keywords -

다국어 초록 (Multilingual Abstract)

This paper explores the potential of Generative AI in the field of interior architecture, with a specific focus on implementing fine-tuning models based on various interior design styles. It highlights that spaces have unique preferences influenced by culture, region, and users, leading to evolving design styles. However, the base models of image-generation AI do not always reflect the changes and latest trends in interior design styles. Model fine-tuning of image-generation AI enables the visualization of spaces incorporating various interior design styles. The study evaluates the base model's performance on 25 diverse design styles, selecting styles for further training based on the results. This fine-tuning involves three main steps: data preparation and preprocessing, text alignment and hyperparameter optimization, and model training and construction. The findings show that fine-tuning effectively represents styles missed by the base model, generating high-quality images. It accurately portrays style characteristics and keywords, offering versatile design possibilities. This research enhances space visualization while accommodating diverse interior design styles and user preferences. It provides a practical mechanism for generating alternative designs to facilitate practical comparisons in interior architecture. Additionally, it highlights the potential application of Generative AI in various fields beyond interior architecture.