Building Rome in a Day: Crowdsourced 3D Cities
Reconstructing Rome, or what 496 computer cores and an autistic savant have in common.
By Maria Popova
Crowdsourcing has clearly been the cultural darling of late. And while some of its most successful applications, from Wikipedia to reCAPTCHA, rely on “active crowdsourcing” — building the collective product by actively soliciting user input — others are starting to work wonders with “passive crowdsourcing,” using user-contributed content that is already available on the web.
Building Rome in a Day, a new project out of the University of Washington GRAIL lab, does just that, using 150,000 Flickr images tagged “Rome” to reconstruct the iconic city in 3D. Because tourist photos are taken from a multitude of vantage points, stitching them together into a cohesive 3D image creates rich and spatially accurate models.
The Rome project, which took 21 hours on a cluster of 496 computer cores, reconstructed some of the city’s most famous landmarks — the Colosseum, St. Peter’s Cathedral, Trevi Fountain and the Pantheon.
The team has also begun reconstructing other cities — Venice, which took 250,000 images and 65 hours, and Dubrovnik, at a more modest 57,845 photos and 22.5 hours.
The ongoing project has fascinating applications in reconstructing not only static landscapes, but also dynamic events as they unfold — in the era of citizen journalism, imagine using public images of anything from natural disasters (like the 2005 Indonesian tsunami) to political protests (like the recent unrest in Iran) to create an accurate record of history.
Or, you can always have a genius autistic savant fly over in a helicopter and draw frighteningly accurate panoramas on a 16-foot canvas.
Published July 31, 2009