Data-Driven Cities and the Use of City Imagery for Improving Accessibility

Cities are the loci of resource consumption, economic activity, and innovation; they are where our looming sustainability problems were born and where those problems must be solved. Given our increasing ability to collect, transmit, store, and analyze data, we have the opportunity to go beyond today’s imperfect and often anecdotal understanding of cities to enable better operations, better planning, and better policy. To understand cities, we must analyze the data exhaust from its components – infrastructure, environment, and people – and how they interact over space and time. While there are already troves of open data about cities, their potential remains largely untapped due to unique challenges related to diversity, scale, and complexity of both urban data and computations required to obtain insights from these data. I will give an overview of methods and tools we have developed for urban data exploration, including our work on the Sounds of New York City (SONYC) project, then present recent work on using large collections of city imagery to study pedestrian infrastructure with the goal of improving accessibility in cities. I will describe Urban Mosaic, a system for browsing large collections of city images; CitySurfaces, an active learning-based framework that leverages computer vision techniques for classifying sidewalk materials; and the NYU-VPR dataset, aimed at helping the development of assistive navigation for the visually impaired population.