DescriptionWhile cities worldwide are increasingly promoting streets and public spaces that prioritize pedestrians over vehicles, significant data gaps have made mapping, analysis, and assessment of pedestrian infrastructure, challenging to carry out. Even in industrialized economies, most cities still lack information about the location, connectivity, and quality of their sidewalks, making it difficult to implement research on pedestrian infrastructure and holding the technology industry back from developing accurate, location-based Apps for different users. Moreover, despite the growing attention to urban data analysis, there is a gap between the real needs of researchers and practitioners directly studying urban problems and the urban analysis tools being developed. Standing at the intersection of economics, urban planning, and computer science, my dissertation aims at addressing both issues by providing theory-rich tools for large-scale assessment of urban sidewalks at two scales: at the human scale, using street-level images by proposing CitySurfaces for classifying eight classes of surface materials, and at the city, scale using aerial imagery, by proposing Tile2Net to create pedestrian networks from aerial imagery. Both studies use computer vision techniques to design frameworks and models for analyzing pedestrian facilities.
This dissertation addresses some of the challenges of semantic segmentation models regarding the high cost of image annotation by employing different techniques, such as active learning to offer solutions tailored to the specific qualities of urban problems.