Revamping Interior Design Workflow Through Generative Artificial Intelligence.

This paper presents some preliminary results on how generative AI (Artificial Intelligence) can transform key tasks in a typical interior design workflow, namely, ideation, schematic drafting, and layout planning. We show via examples how targeted fine-tuning and deliberated prompt engineering can jumpstart the creative process and improve the overall workflow efficiency and effectiveness. Ideation by collecting targeted reference designs to fine-tune a Stable Diffusion model and cranking out various ideas by simply engineering prompts. Schematic drafting is to test design ideas in the actual environment, where designers model the space structure with specific design elements. We used an actual photo of a plain office as a guidance and teach the model to generate images with specified features, such as different styles, materials, décor objects, while maintaining the underlying spatial configuration. Layout Planning is to organize the functions with comprehensive consideration of functional relationships and layout design principles. The generative model is fine-tuned using selected layout plans, which can accommodate different functional strategies, such as private rooms, large open areas, etc. and provide viable layout options.