01 June 2007


Like a lot of people, I've been charmed by Google Maps Street View, which allows you to dive into a map to see panoramic photographs taken at street level. Google drove trucks around cities like my own San Francisco, took a bunch of photographs, and linked them to their existing map data. Beautiful, clever, useful, and exactly the kind of enriching-the-world-of-available-data thing Google is there for.

It got me thinking of that NASA composite image of the Earth without cloud cover which was made by digitally combining many, many images of the Earth from space. Gee, I thought, wouldn't it be great if Google were to drive those trucks around several times, and then do a digital composite that hid all of the cars and pedestrians?

It turns out that Google doesn't need to send the truck around again to do this. The technology already exists to scan a big image archive—say, flickr—and then see where the pictures overlap and develop a 3D model of the space that has been photographed.

Follow that link if you have any love of these kinds of things; the demo is amazing.

So before long, having linked some reference photos to maps, Google should be able to produce a much more detailed version of Street View based on the collective photographic knowledge of the entire web.

This idea had me believing in the Singularity for a good fifteen minutes.

No comments: