Deep Learning the City : Quantifying Urban Perception At A Global Scale
This work addresses the problem of limited throughput in urban perception studies for researchers and planners, though it is incremental as it builds on existing computer vision methods with new data and a hybrid approach.
The authors tackled the challenge of quantifying urban perception globally by introducing a large crowdsourced dataset of images and pairwise comparisons, and training a Siamese-like CNN to predict human judgments, achieving scalable urban perception data.
Computer vision methods that quantify the perception of urban environment are increasingly being used to study the relationship between a city's physical appearance and the behavior and health of its residents. Yet, the throughput of current methods is too limited to quantify the perception of cities across the world. To tackle this challenge, we introduce a new crowdsourced dataset containing 110,988 images from 56 cities, and 1,170,000 pairwise comparisons provided by 81,630 online volunteers along six perceptual attributes: safe, lively, boring, wealthy, depressing, and beautiful. Using this data, we train a Siamese-like convolutional neural architecture, which learns from a joint classification and ranking loss, to predict human judgments of pairwise image comparisons. Our results show that crowdsourcing combined with neural networks can produce urban perception data at the global scale.