Revisiting Gehl through perceived safety: A computational method

This study examines how ground-floor urban form relates to perceived safety in Stockholm, Sweden, using approximately one million street view images collected between 2010 and 2021 together with multi-source built-environment data. Jan Gehl’s writings discuss perceived safety as part of a human-centered design framework that integrates both traffic- and crime-related safety through street-level environmental design, drawing together ideas also established in broader planning and safety scholarship. Building on Gehl’s integrative framework, we operationalize a set of design attributes emphasized in his work and evaluate their associations with perceived safety scores derived from locally collected image-comparison surveys and extended citywide using deep learning models. Across multiple modeling approaches and two spatial scales, ground-floor features related to transparency and public–private transitions show the strongest and most consistent positive associations with perceived safety, often exceeding the explanatory strength of neighborhood demographic and land-use controls. Other features such as pedestrian and cycling infrastructure also demonstrate meaningful relationships, particularly at a finer spatial scale. These findings suggest that perceived safety is closely associated with the human-scale, street-level design features integrated within Gehl’s framework and demonstrate how computational urban analytics can operationalize qualitative design principles and evaluate their relevance to urban livability at scale.