Researchers Create Picture Dictionary for Cities
Researchers used a vast library of Google Street View images to find the visual features that define some of our greatest cities.
What gives your city its unique visual character?
This may sound like a pretty subjective question, but researchers from Carnegie Mellon and a Parisian university are making it less so. They’ve developed visual data mining software to detect which features define a city’s look.
For Paris, the team concluded that cast-iron balconies, windows with railings, special Parisian lampposts, and certain styles of doors make up the city's unique identity.
For New York it's the fire escapes, and bay windows defined San Francisco, though the software had trouble picking up on elements in U.S. cities because of a relative lack of stylistic coherence compared to European cities. Look at more of the findings here.
The software surveyed more than 250 million visual characteristics from 40,000 Google Street View images of Paris, London, New York, Barcelona, and eight other cities.
Then a machine learning program analyzed the images and determined which details differentiated the images from ones taken in other cities. The process requires enough computing to keep 150 processors working overnight.
The team of researchers from CMU and INRIA/Ecole Normale Superiéure will present their findings on August 9 at the SIGGRAPH 2012 convention at the Los Angeles Convention Center in Downtown Los Angeles.
So what features define your city? Will the new system help cities find their identity, or is it just telling us what we already know?
Photo via CMU