Abstract Summary
While leveraging urban streetscape data and deep learning is highly effective for identifying general patterns and analyzing large areas in urban public space studies, its utility is somewhat limited when seeking specific design recommendations. This approach tends to overlook smaller yet impactful elements, such as wires and seating, due to their minimal visual area in streetscape data. This limitation constrains the depth of insights into how all elements influence users' spatial perceptions in urban public spaces, particularly in identifying unrepresented elements or gathering users' needs for spatial interventions. In response to this challenge, our study focused on Beijing's public spaces to discern which urban elements impact users' spatial feelings and how adjustments to these elements could foster more positive urban public environments. Through field and online surveys across 32 public spaces, involving feedback from approximately 2000 respondents, and employing computer vision and natural language processing for data analysis, we established a pathway from user perceptions to actionable urban design interventions. This method aims to offer an "end-to-end" strategy for improving urban public spaces by transitioning from user experiences to the implementation of designs that cater to user needs. Although our initial findings call for further statistical scrutiny and broader empirical validation, our objective is to develop a more integrated approach to urban survey and analysis that can directly inform urban planners and designers, reducing reliance on extensive datasets and advanced technical tools.
