Sidewalk Ballet: A scalable urban analytic approach for explaining social qualities of urban streets

About

Why do some streets attract more social activities than others? Is it the design of the street, or is it the land use pattern within neighborhoods that allows businesses around which people congregate and meet each other? Or is it instead the demographic characteristics of the residents that make some areas more prone to neighborly interactions than others? This research investigates why some urban streets feel lively and inviting while others don’t, analyzing how street layouts, land use, and local communities shape social vibrancy. Using AI and street-level imagery from sources like Google Street View and Apple Look Around, the team developed a system that detects people in public spaces and identifies whether they’re interacting—creating the first large dataset focused on urban social behavior.

The project’s findings help urban planners and policymakers design more inclusive, walkable neighborhoods that promote community connections and public health. It also supports sustainability goals by encouraging pedestrian-friendly spaces and uncovering how infrastructure gaps may limit social opportunities in marginalized areas. Key achievements include an advanced AI model for spotting social interactions, a balanced dataset of over 70,000 annotated examples, and a expanded library of street imagery for broader analysis.