Innovative Design Method for Lingnan Region Veranda Architectural Heritage (Qi-Lou) Facades Based on Computer Vision

The veranda buildings (Qi-lou) in Lingnan are important historical and cultural carriers of knowledge of the transformation of cities in Lingnan from traditional to modern forms. However, the restoration and protection of Qi-lou building facades faces challenges such as low efficiency, over-reliance on architects’ experience, and a lack of systematic technical tools. This study adopts computer vision techniques, specifically conditional generative adversarial networks (CGANs), to propose a new framework for generating realistic Qi-lou building facade designs. By collecting and processing 131 Qi-lou building facade samples, this study constructs a database covering various architectural styles, such as Gothic, Baroque, and Nanyang styles, and introduces a building exterior profile (BEP), a functional segmentation layout (FSL), and a three-stage hierarchical modeling approach for building elevation (BE) to achieve stylized facade generation. The main research findings are as follows: (1) The proposed CGAN model can accurately generate functional zoning layouts and effectively improve the design efficiency of Qi-lou building facades. (2) The generated facade images are consistent in style and structure, exhibiting high realism and significantly reducing the need for manual adjustments. (3) The generation quality of the model is significantly optimized as the input complexity (such as the number of windows and colonnades) increases. This shows that the framework proposed in this study can provide effective technical support for the restoration and innovative design of Qi-lou buildings in the Lingnan region.

This creative work has been published under the Creative Commons Attribution 4.0 International (CC-BY 4.0) license which allows copying, and redistribution as well as adaptation of the original work provided appropriate credit is given to the original author and the conditions of the license are met.