Measuring the Multiple Functions and Tradeoffs among Streets: A New Framework Using the Deep Learning Method

Abstract

With the sustainable and coordinated development of cities, the formulation of urban street policies requires multiangle analysis. In regard to the existing street research, a large number of studies have focused on specific landscapes or accessibility of streets, and there is a lack of research on the multiple functions of streets. Recent advances in sensor technology and digitization have produced a wealth of data and methods. Thus, we may comprehensively understand streets in a less labor-intensive way, not just single street functions. This paper defines an index system of the multiple functions of urban streets and proposes a framework for multifunctional street measurement. Via the application of deep learning to Baidu Street View (BSV) imagery, we generate three functions, namely, landscape, traffic, and economic functions. The results indicate that street facilities and features are suitably identified. According to the multifunctional perspective, this paper further classifies urban streets into multifunctional categories and provides targeted policy recommendations for urban street planning. There exist correlations among the various street functions, and the correlation between the street landscape and economic functions is highly significant. This framework can be widely applied in other countries and cities to better understand street differences in various cities.