| Follow Us:
Tressa Glover and Don DiGiulio of No Name Players.  Photograph by Brian Cohen
Tressa Glover and Don DiGiulio of No Name Players. Photograph by Brian Cohen | Show Photo

Innovation

What makes Paris look like Paris? Software pinpoints elements that give cities distinctive flair

What makes Paris look like Paris? 
 
CMU has created software that pinpoints the visual elements that give certain cities their distinctive flair, such as the doorways of Dublin or the balustrades and bridges of Paris.
 
Of course you don’t need a computational data tool to see that places such as Paris and Barcelona have a look that sets them apart from other cities. But this research gives us visual data mining software that scientifically confirms this fact while identifying the subtle features that are unique to each place, such as the cast-iron balconies in Paris, fire escapes in New York City and bay windows in San Francisco.
 
Researchers analyzed more than 250 million visual elements from Google Street View in 12 major cities in the world to crack their stylistic code. Paris might be the most harmonious of cities studied, from its lampposts to doors, balconies, windows with railings and the shape of its street signs.
 
U.S. cities, not surprisingly, showed a relative lack of stylistic coherence, no doubt a result of their melting pot of styles, researchers say.
 
This is the first time that finding patterns in large databases--called “Big Data mining”—has been used to identify visual elements, says Alexei Efros, associate professor of robotics and computer science. It may one day prove useful for computational geography tasks.

The research was conducted by CMU and INRIA/Ecole Normale Supérieure in Paris.
 
In other news, CMU researchers report they have mastered a streamlined approach to the painstaking technology behind computer animated movies and games as part of their work for Disney Research, Pittsburgh.
 
Computer graphic artists often spend hours creating an animated character’s subtle movements—such as a yawn or the drape of their clothing. Improving the method of modeling these movements will greatly simplify the editing process and enable artists to create more compact, powerful and easy-to-manage models. 
 
Both findings will be presented on Aug. 6 at SIGGRAPH 2012, the International Conference on Computer Graphics and Interactive Techniques at the Los Angeles Convention Center. 
 
 
Writer: Deb Smit
Source: CMU
 
Signup for Email Alerts
Share this page
0
Email
Print
Signup for Email Alerts