An Enhanced Method of Interior Design Generation Based on Stable Diffusion Model with Hand-Drawn Sketch

The application of generative artificial intelligence in interior design is gradually becoming a new research direction in the field of interior architectural design. Due to limitations in the content richness, quality, and controllability of interior design images, current related work is still in the exploratory stage. To address this, this paper proposes an interior design text-to-image method that combines Stable Diffusion with hand-drawn sketches of interior design. This paper initially constructs the Interior Design Dataset (IDD), followed by fine-tuning the Stable Diffusion model on this dataset using the LoRA (Low-Rank Adaptation) technique. Subsequently, the ControlNet model is employed in combination with hand-drawn sketches of interior design to achieve further control over the output. Finally, extensive experiments are conducted in this paper, and comparisons are made with current representative text-to-image models, including PaddlePaddle AI Studio and DALLE-3. The results demonstrate that the proposed method outperforms the others, enabling the efficient generation of various interior designs, promoting the creation of innovative designs, and enhancing both design and decision-making efficiency.