Reimagining Residential Planning: A New Paradigm Using Grasshopper Combined with Fine-Tuned Diffusion Models

Abstract

Rapid global urbanization has increased the demand for residential construction, rendering traditional design methods insufficient. This study proposes an innovative approach integrating parametric techniques with fine-tuning diffusion models to automate the modeling and rendering of residential area planning. Specifically, this method first utilizes Grasshopper programming to develop a parametric design system that can automatically create 3D models of residential areas. Subsequently, a new dataset was constructed to fine-tune the diffusion model, enabling it to generate high-quality renderings that conform to specific architectural styles, types, and viewpoints. The proposed method transforms the traditional scheme refinement, modeling, and rendering processes into a generative design process. In a subjective evaluation of 2880 generated images by ten designers, the proposed method obtained higher scores than Midjourney, DALL E3, and Stable Diffusion in usability (81%), building style consistency (87%), and realism (92%). Results indicate that this method can efficiently generate diverse design options at the preliminary design stage. This study provides a practical workflow for the rapid generation of residential area models and renderings.