Large-scale measurement of street interface transparency based on street view data and deep learning: A study of first-tier cities in China (Beijing, Shanghai, and Guangzhou)

With the current transformation toward more refined urban planning and design, the study of street space quality from a human-scale perspective has increasingly gained widespread attention. As one of the key factors influencing street space quality and vitality, street transparency—defined as the proportion of ground-level openings (windows and doors) in the total ground-level facade area—has seen a growing demand for quantitative measurement. Traditional methods of measuring transparency rely mainly on costly and inefficient manual analysis, making large-scale, efficient measurement difficult. To address this issue, this study proposes a set of methods for large-scale and detailed measurement and analysis of street transparency from a human-scale perspective, based on open-source street view images and deep learning. Using the central urban areas of first-tier cities in China (Beijing, Shanghai, and Guangzhou) as examples, this study successfully and efficiently calculates and compares street transparency in these areas and analyzes both macro- and micro-level factors influencing transparency and reveal specific characteristics and development patterns of transparency rate. The study finds that street transparency in the central urban areas of these cities exhibits significant spatial heterogeneity, showing a spatial pattern of “higher inside and lower outside,” and identifies the penetration rate is not only affected by macro factors such as economic construction and commerce but also strongly affected by micro factors such as the design of the walkway and the layout of the street on both sides of the street, providing a basis for targeted policy guidance and institutional design in urban planning.