The Technium

Crowdsourced 3D City

If a million monkeys typing can write Hamlet, can a million tourists’ snapshots map out a 3D picture of Rome?

The answer is yes. And Venice, too.

There are more than 2 million photos on Flickr tagged with Rome. They capture almost every nook and cranny, every column and doorway, of the old city. If you had a lot of computer power, and the right smart software, you could take these 2 million views and compile them into a single unified 3D portrait.

And if this system from the University of Washington GRAIL lab worked for Rome you could take other highly photographed places, such as St Marks Plaza in Venice, and construct a visual 3D picture, down to the smallest detail.

Picture 10

Microsoft’s Photosynth accomplished a similar feat. The movie below does that same thing with huge piles of photos taken off of  Flickr.

Eventually, the every city in the world will get a full textured 3D view of itself. (via Waxy)

  • William Pietri


    The obvious next stage, and one I have an urge to start work on immediately, is a game where people fill in the missing pictures.

    Your smartphone could drag you to the right spot and show you the right angles. Then you could shoot low-res phone photos for some points, or high-res photos for more points. When places change, you’d get points for being the first to notice and photograph. And your phone could remind you as you pass of areas where photos are stale, encouraging you to keep an eye out.

  • frank mcgillicuddy

    how do you avoid time decay?? I presume you have the date of the photo, so you should then have a series of time-shifted versions of the locations, and the version of the summation of the photos should have a date-range displayed – i wouldn’t want to mix stuff from too vastly different dates, and ofcourse cities are always revamping/upgrading/repairing themselves…..

    neato mosquito

  • Bernie

    I recommend you don’t avoid time decay but embrace it. Not only can you see the transformations, but you can reverse them or cross reference them with other data you have.

    Such photos provide not only a spatial reference, but a temporal one.

  • @William: Wow, that’s a great idea. Some way to get “assignments” about the undone parts that are still needed. In fact, it could be a of kind of treasure hunt to find areas not yet photographed!

  • james

    I’ve been speaking to colleagues recently about how we use this technology for 4d modelling – an extension of the temporal reference Bernie refers to.

    Take a location, much photographed since the dawn of photography – London, UK or Paris, France or New York, NY and start building a city you can explore in both space and time.

    The photos need to be time and geotagged, and there will be some vagueness initially and dispute continuously. Some of this work might be outsourced though in the way historical collections are already being added to flickr.

    So then you go “show me St Paul’s Cathedral in 1939″, move around it to find an angle you like, then “take me through to 1946″.

    Where data for a time period is missing, ghost an image from the nearest time period, less opaque the further distant it is. (B&W then sepia going backwards for images taken previously, some other signified for images taken subsequently, what?)

    Start adding frames from moving images 1890s onwards.

    Add processing power, then start throwing random untagged (or poorly tagged) images at it – “what is this?” Where is this?” “when is this?”