Using 50 million Google Street View images of cars in 200 American cities, the study's researchers determined, with the help of a "machine vision framework based on deep learning," the make, model, and year of each car (2,657 categories). They then used that information to "accurately estimate income, race, education, and voting patterns, with single-precinct resolution." The average precinct has a population of only about 1,000, say the researchers. Here are some of their findings, in their own words...
- "We successfully detected 22 million distinct vehicles, comprising 32% of all the vehicles in the 200 cities we studied, and 8% of all vehicles in the United States."
- "Our model detects strong associations between vehicle distribution and disparate socioeconomic trends."
- "The vehicular feature that was most strongly associated with Democratic precincts was sedans, whereas Republican precincts were most strongly associated with extended-cab pickup trucks."
- "Our estimates accurately determined that Seattle, Washington is 69% Caucasian."
- "We estimated educational background in Milwaukee, Wisconsin zip codes, accurately determining the fraction of the population with less than a high school degree."
The researchers ask whether this type of analysis eventually could replace costly and time-consuming door-to-door efforts such as the American Community Survey. "As digital imagery becomes ubiquitous and machine vision techniques improve, automated data analysis may provide a cheaper and faster alternative," they suggest.
Source: arXiv, Using Deep Learning and Google Street View to Estimate the Demographic Makeup of the US