This comment republishing feature is annoying -- I do not give anyone permission to copy my work here to other sites. The use of the Y Combinator logo is also suspect; I do not speak for YC when I post here (or ever), so I am not sure why my comments would have a YC logo on them.
Anyway, commenting on the actual content... I guess the algorithm builds 3D models of structures from the photo. Initially, I thought it would generate something like Street View. It seems possible to use the 3d dot model in addition to the photos to create a "photorealistic" virtual world -- has anyone done this? (This article is very low on details.)
Its very similar to what you can do with Microsofts Photosynth, just on a greater scale. MS lets you upload a couple hundred pictures, which is enough for a similar 3D dot model for a single building.
Actually, this is Photosynth (rather both use th same base tech.) In fact, city-wide synthing has been possible for some time now- the crux is matching up the photos fast enough, which has now been reduced to a day (for 1M photos). For the interested, fast (real-time) matching of video streams was also demoed at TechFest some year or so ago. I'm pretty sure there's a video about it somewhere, possibly at TechCrunch.
According to the article "the team previously developed tools which can create 3D models from a collection of photos, which subsequently evolved into Microsoft’s Photosynth." I can not wait to see the evolution of this stuff, especially from Bing and Google maps, in the coming years.
I am very impressed with this project. In my opinion this is even better than Google Maps, which relies on hand modeled structures. I'd like to see Google implement Flickr reconstructions so that even if no one has hand modeled buildings there will still be 3D structures there when you zoom in.
Naturally not every place will have such a high photo density but there must be a lot of cities and tourist sites that can be reconstructed in a similar manner.
Something like that was announced just yesterday, or very recently, Google is going to use traffic camera, security camera footage to liven up their maps with realistic crowd and vehicles, plus I assume part of they could use any building photo too for their 3D map.
Anyway, commenting on the actual content... I guess the algorithm builds 3D models of structures from the photo. Initially, I thought it would generate something like Street View. It seems possible to use the 3d dot model in addition to the photos to create a "photorealistic" virtual world -- has anyone done this? (This article is very low on details.)