Using Open Data and Deep Learning to Explore Walkability: A Big Data-Driven Study

Keywords: Walkability, Street view images, Deep learning, Big data
Abstract Type: Virtual Paper Abstract

Abstract

A measure of how the built environment is pedestrian friendly is walkability. The increasing availability of open data and advancement in artificial intelligence in recent years provides us with a great opportunity to develop an innovative methodology to study walkability and plan walkable cities. This paper proposes a novel walkability framework which quantifies walkability from four aspects: safety, convenience, continuity, and attractiveness. Using Shenzhen as our case study, we extract built environment features from different data sources. Deep learning semantic segmentation approaches are applied to identify street elements from street view images. Results show that urban areas and suburban CBDs have higher walkability than other areas, and the majority of the areas have poor walkability scores. Our findings uncover the limitations of the built environment for walking, and provide implications for stakeholders and practitioners.