Can You Tell a City By Its Windows? Your Computer Can

• August 20, 2012
• 1 Comment

by Shilpi Paul

Courtesy of Carnegie Mellon.

Is a city’s personality more apparent in its landmarks or its doorknobs? A new program from Carnegie Mellon may provide the answer.

Researchers at the Pittsburgh university have devised a program that scans through 40,000 Google Street View images from 12 cities to try to find the personality of a city through the infinite collection of street scenes, buildings and sidewalks, reported TechCrunch this weekend.

The program finds clusters of identifying images that are common in a particular city, but not necessarily common in others, like particular trees in Barcelona, lamp posts in Paris, cast iron railings in London and bay windows in San Francisco. The researchers concluded that rather than by one landmark, like the Eiffel Tower or the Capitol, cities are identifiable by these sorts of ubiquitous characteristics.

Courtesy of the report

One very intriguing finding is that European cities officially have a more unified look than U.S. metropolises: the algorithm didn’t work too well on most American cities, and the computer had a much harder time finding common characteristic features. Perhaps the heterogeneity is due to waves of development that bring in entirely new architectural styles at different times, or simply our diverse tastes as a society. (The report cited the “relative lack of stylistic coherence and uniqueness” — ouch).

At the very least, the historic preservationists now have one more tool at their disposal when they want to figure out what makes a city, a city. Here is the full report from Carnegie Mellon.

This article originally published at https://dc.urbanturf.com/articles/blog/can_you_tell_a_city_by_its_windows_your_computer_can/5927

1 Comment

1. xmal said at 4:08 pm on Monday August 20, 2012:
   Cool find!