Research on the Perception of Community Public Space Vitality and Renewal Strategies Based on Deep Learning and Street View Data

Abstract

Community public spaces serve as essential venues for the daily activities of urban residents, and their level of vitality directly affects the quality of the living environment and the sustainability of urban development. To scientifically evaluate the vitality of community public spaces and propose targeted optimization strategies, this study integrates deep learning techniques with multi-source street view data to construct a vitality perception evaluation framework focusing on five core dimensions: aesthetics, recognizability, functionality, comfort, and safety. Taking two representative neighborhoods in Shanghai—one in the city center and one in the suburban area—as case studies, the research employs visual semantic segmentation to perform pixel-level analysis of street view samples, quantifying the proportions of spatial elements such as buildings, greenery, roads, and public facilities. These data are then combined with Point of Interest (POI) information for kernel density and diversity analyses. The results show that central urban neighborhoods perform better in pedestrian paving, street furniture, POI density, and road accessibility, indicating a higher overall vitality perception level. In contrast, suburban neighborhoods exhibit slight advantages in traffic signage but generally suffer from lower green visibility, encroached pedestrian spaces, and insufficient green and environmental facilities, which constrain spatial vitality. The findings suggest that both types of neighborhoods insufficiently address user experience. Accordingly, this study proposes four renewal strategies—enhancing walkability, improving environmental attractiveness, enriching functional services, and strengthening safety perception—to maximize the effectiveness of public space use and provide a scientific basis for the optimization of urban living environments.