Impact of visual features and urban form on street vitality using explainable AI: a study of historic districts in Shanghai

Street vitality in historic districts serves as a critical indicator of sustainable urban development. While morphological factors – such as street networks, building characteristics, and land use – are recognised as influential, visual features play a distinctive role in shaping pedestrians’ perceptions and spatial preferences, particularly in historical environments. However, the combined effects and interactions among these variables remain underexplored. This study investigates how street network configuration, architectural features, and visual elements jointly influence street vitality in historic central districts. We integrated spatial data with 45,745 street-view images to extract visual features and estimate pedestrian flow. Notably, this research introduces an innovative vision–3D urban form framework that enhances the precision of quantitative analysis. Using a LightGBM regression model, we quantified the relative importance of each variable. Results indicate that the number of intersections (9.0%), parking facilities (7.4%), and shop density (7.1%) positively contribute to pedestrian activity. Building height (5.5%) and height variability (3.7%) influence attractiveness in nonlinear ways. Visual elements – including sky-view ratio (3.3%), ground surface ratio (3.0%), colour saturation (2.9%), and brightness (2.0%) – also contribute to street appeal, with vivid colours associated with higher pedestrian volumes. The study highlights threshold effects and offers data-driven insights for micro-renewal strategies that support more pedestrian-friendly historic urban spaces.