This is a paper being submitted to SIGGRAPH (a big computer graphics conference) describing a method of describing a 3d scene on a picture to a computer so it can render in a 3d object and make it look real, with physics and lighting and everything.

http://kevinkarsch.com/publications/sa11.html
examples:


There's an embedded video in the link that demonstrates how amazing the tech is.

That's pretty cool, I could see where it would have so many uses. The first question when I looked at the pictures was "It's all well and good having someone insert and object affected by the existing light-sources but what about adding light?" then low and behold, the first thing it showed was a spherical light-source being added. The results were very good! I mean, you can see the odd flaw, but that's to be expected.

I guess if this gets properly released, the next stage is the ability for the program to recognise more in the original photo without the user interaction.

Nice stuff.

Yea, it's cool. I'd like to see this kind of thing applied to making game levels. Take a few pictures around the house, spend 30 minutes defining walls and boundaries and crap, then presto, your house is a COD level or something. That would be amazing... and really save in development time for some stuff.