What's Your Street's Personality?

Researchers at Carnegie Mellon University in Pittsburgh, USA have devised an algorithm to pick up on various elements that contribute to the character of streets using Google Streetview for data.

According to this 'american city' article, "using... Google Street View, the researchers developed an algorithm that detects elements, such as a window, column or balcony, that are both distinct and occur with regularity inside a city"

The software relies on elements that are frequent and distinct to decide what is representative of an area.

" Noting more than 250 million features in a dozen well-photographed cities, the researchers compiled a database that allows them to pick up on patterns that form a city’s architectural fabric"

The website also contains a neat video explaining the software, using it to define what characteristics make Paris, Paris.

I like how this software can pick up on these particular elements and use them for cross referencing. This allows for comparisons at various scales; street, city, country even at a continental scale (as shown in the  video ).  I wonder what the thresholds are - how often does an element need to be repeated to be representative?  I'd love to test out this software, if accurate (and sensitive) enough, it could potentially assist planners with neighbourhood character studies, otherwise at least it could provide an interesting tool to look at common architectural elements in your neighbourhood.

Read on here.  Click here to see the video.