Replacing Google Street View with complete 3D models of cities [video]

The problem with Google Street View and Maps is that with rare exceptions, its data is entirely two dimensional. Satellites look down, after all, while Street View has captured only a very rough first draft of the world’s streets.

That’s where the work of Noah Snavely, a computer scientist at Cornell, comes in. You may not know his name, but you’ve probably seen his work in the form of Microsoft Photosynth, which is based on his code.

Snavely has pulled a number of tricks since his initial PhD work in the mid-aughts, from 3D visualizations of the Statue of Liberty based solely on tourist’s photos to a point cloud realization of the Colosseum in Rome. (You can read about his work in detail in an accessible paper (pdf) on the subject published in 2010.)

One of the problems Snavely faces in getting sufficient coverage of cities is that reconstructing a three dimensional model of a street requires many more pictures than are captured by, for example, the cameras of Google Street View.

That’s where the online game PhotoCity comes into play. The idea is to incentivize people to take pictures of particular spots on buildings, even their own neighborhood.

Another way to fully map cities in three dimensions is simply to get better data out of the photos that are already online. At this year’s Emerging Technologies conference at MIT, Snavely said that if he had one wish, it would be that all cameras would accurately record the place where a user took a picture. Given that smart phones, which have build in location services, are now poised to capture more photos than conventional digital cameras, it might not be long until Snavely gets his wish.