Abstract
Understanding the socioeconomic and demographic characteristics of an urban region is vital for policy-making, urban management, and urban planning. Auditing socioeconomic and demographic patterns traditionally entails producing a large portion of data by human-participant surveys, which are usually costly and time consuming. Even with newly developed computational methods, amenity characteristics such as typeface, color, and graphic element choices are still missing at the city scale. However, they have a huge impact on personalized preferences. Currently, researchers tend to use large-scale street view imagery to uncover physical and socioeconomic patterns. In this research, we first propose a framework that uses deep convolutional neural network to recognize the typeface from street view imagery in London. Second, we analyze the relationship between 11 typefaces and the average household income in 77 wards of London. The result show that the typefaces used in the neighborhood are highly correlated with economic and demographic factors. Typeface could be an alternative metric to evaluate economic and demographic status in large-scale urban regions. More generally, typeface can also act as a key visual characteristic of a city.