Tired of the faceless urbanites crowding their Google Street Views, computer scientists aimed to remove the pedestrians entirely. The images above show they succeeded, mostly.
The software was developed by Arturo Flores of the University of California, San Diego; earlier this summer he unveiled (pdf) the proof-of-concept. It’s built off of a previous algorithm developed in 2005 that can pick out pedestrians in urban settings. The new program removes the identified pedestrian and covers the gap using pixels from slightly ahead and slightly behind what appears to be someone walking down the street. But it only works in cities (where tall buildings give a relatively flat backdrop), can create a human smear when the photographed person walked at the same speed as the Google camera, and, one could imagine, has trouble in huge crowds–where neighboring pixel-swapping might result in blurry Frankenhumans.
But otherwise, it leaves a non-distracting, relatively “ghost free” image, a university press release says, that will further protect pedestrian privacy. When it almost succeeds, it gives users a good laugh: a post-apocalyptic cityscape including disembodied feet, ownerless dogs, and floating umbrellas.
Related content:
Discoblog: Confused (and Injured) Pedestrian Sues Google Maps Over Bad Directions
Discoblog: And the Survey Says: Google Is Not Making You Stupid
Discoblog: My Name Is Topeka, Kansas, but You Can Call Me Google
Discoblog: Tweet Your Prayers, Google Your Ancient Texts
Images: Arturo Flores
