Research on a ControlNet-based facade generation methodology for rural self-built houses with the goal of optimizing government-issued general-purpose design atlases

In response to the rapid urbanization of China and the surge in demand for rural self-built housing, the Chinese government has implemented a policy of government-issued general-purpose design atlas for rural self-built housing. However, this policy encounters problems such as insufficient supply of designs, overly specialized drawings, and inconsistent design styles. To address these issues, herein, we collect 908 rural self-built housing design schemes in Zhejiang Province. Following expert screening, standardization, mask and map drawing as well as prompt engineering, multimodal generation technology based on Stable Diffusion and ControlNet is employed to conduct hierarchical model training for mask drawing, map drawing, and rendering, thereby developing a preprocessor capable of generating customized renderings of rural self-built housing exteriors. The preprocessor utilizes architectural exterior masks and prompts as conditional information to generate architectural renderings that accurately describe the architectural exterior component system as well as its colors and materials. This technology offers a forward auxiliary design for rural self-built housing from scratch, enabling rapid generation and iteration of designs, diversifying the selection of existing general-purpose design, making the results more visual, and addressing the problem of architectural style harmonization and unification in rural areas to a certain extent.