In an era where extreme amounts of data are produced daily, with the majority being visual content, cities have become vast repositories of digital imagery. The book traces the evolution of visual analysis in urban studies, from early photographic documentation to how today's sophisticated AI is revolutionizing our understanding of cities.
A walk through Cambridge, Massachusetts, shows how much a city's character, history, and inequality can be read from what's simply in view — and why one perspective is never enough. The introduction frames the book's central subject, visual artificial intelligence, and the ethical questions that come with it.
From medieval guidebooks to William Whyte's street-corner films and Kevin Lynch's mental maps, scholars have long tried to systematically observe cities. This chapter traces that lineage into the present, where computational models sort through an overwhelming volume of urban imagery — and asks what biases they might inherit.
Digital images aren't just pictures; they're arrays of numbers, and that changes everything. This chapter follows the leap from analog photography to smartphone cameras and street view imagery, and introduces the techniques, from segmentation to neural networks, that make large-scale visual analysis of cities possible.
How do you teach a computer to recognize a sidewalk, a storefront, a stranger? This chapter traces the breakthroughs behind modern computer vision, ImageNet to convolutional neural networks and the generative models now used to both analyze and imagine urban space.
What if you could measure a city's greenery, pixel by pixel? This chapter follows that idea from early green-view indices to today's semantic segmentation methods, and the surprising "green bias" that appears depending on whether you measure from the street or from above.
Beyond the street, indoor spaces reveal how we actually live. Drawing on over 640,000 Airbnb photos from 80 cities, this chapter shows that despite a global drift toward sameness, geography and culture still leave a visible fingerprint on how people furnish and arrange their homes.
Safety, beauty, liveliness sound very subjective; but this chapter shows they can be measured. Building on crowdsourced image comparisons, it explores how visual AI, combined with physiological data like eye-tracking, is starting to quantify how urban design actually makes people feel.
With more than a billion CCTV cameras worldwide, this chapter looks squarely at surveillance — from camera-dense cities to million-camera networks — weighing real gains in safety and traffic management against the risks of algorithmic bias and a society that is always being watched.
Generative AI can paint a van Gogh-style Cambridge in seconds — but whose data trained it, and whose biases came along for the ride? This chapter examines generative models in urban design, from playful reinterpretations of literature to their growing use by architects and planners.
Text has always been the city's memory — zoning codes, paths and landmarks — but it rarely captures how a place actually feels. This chapter looks at multimodal AI, which links images, language, and sensor data into a single, more complete way of understanding urban life.
The book closes by confronting what visual AI still can't see — informal settlements invisible to street view and satellite imagery alike — through a case study of LiDAR mapping in a Rio de Janeiro favela, and a case for transparent, community-driven urban data.