Abstract: Much of urban life remains unmapped, limiting our ability to plan at the human scale. My research tackles this gap by designing open-source tools and frameworks that combine computer vision, machine learning, and computational geometry to extract critical urban features from imagery and open data. Moving beyond static maps of where infrastructure exists, my work decodes how streets function and evolve over time. This spans from visual analytics systems that 3D reconstruct and analyze crossing patterns at intersections using video data, to a change retrieval framework that fuses aerial imagery, street geometry, and design records into a unified temporal model. In this talk, I’ll share how these tools can enable a more complete, analyzable, and actionable view of cities: one grounded in how people actually move through and experience urban space.
