Work

A knowledge-augmented diffusion framework for generating regionally distinctive architectural facades: a case study of Huizhou dwellings

Year: 2026

Type: article

Abstract: Generating heritage facade images with recognizable regional style is increasingly feasible with diffusion models, but prompt engineering or standard fine-tuning alone rarely encodes the construction logic, spatial hierarchy, and symbolic rules that make such heritage architecture intelligible. This gap is critical for Huizhou traditional dwellings (HzTD), a distinctive type of vernacular architecture in southern Anhui Province, China, whose facade forms depend on typological relations among courtyard layouts, timber-frame bay rhythms, materials and decorative motifs. To address this gap, this study proposes a Knowledge-Augmented Stable Diffusion Model (KA-SDM) for semantic representation and controlled facade generation of HzTD. The framework first organizes construction knowledge into a CRM-inspired property graph schema implemented in Neo4j, where posts, spaces, ratios, wall plans, elements, and details are represented as queryable entities and relations. Rule-based graph inference then derives facade composition attributes, opening configurations, and detail semantics, which are converted by a Prompt-Conditioned Semantic Controller (PCSC) into structured prompt fragments and ControlNet masks. Experiments on Huizhou dwelling facade images from Xidi Village compare KA-SDM with a Stable Diffusion (SD) baseline and ablation settings based on LoRA and ControlNet. The results indicate that knowledge-guided semantic control improves typological consistency, structural plausibility, and expert-rated semantic alignment while preserving visual diversity. This research demonstrates a practical pathway for coupling architectural heritage knowledge representation with generative AI, supporting cultural documentation, design assistance, and virtual reconstruction.

Source: Architectural Intelligence

DOI: 10.1007/s44223-026-00137-3

Authors: Zhe Guo, Yuhang Kong, Zihuan Zhang, Qingyi Xiong

Institutions: Tianjin University, Hefei University of Technology

Language: English

Topic: Generative Adversarial Networks and Image Synthesis

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: diamond

Funders: National Natural Science Foundation of China, Fundamental Research Funds for Central Universities of the Central South University, Humanities and Social Sciences Youth Foundation, Ministry of Education of the People's Republic of China

Awards: G3794874485, G6491721595, G8637621446