Photographs Converted to 3D models (via www.gizmag.com)
This ranks high up in the cool factor if is true. A company called fotowoosh claims to have developed an algorithm to estimate depth in photographs and to transform said photographs into 3d models.
They have some animations on their site showing walkthrough 'inside' a photo and I must admit that the result is stunning.
The website doesn't state one way or another, but I have a feeling the algorithm is not fully automatic. I believe the user is prompted to select the correct perspective from various 3D outcomes the algorithm generates.
As you know, it takes two eyes -2 2D dimensional images - to create perspective in your brain. Fotowoosh would work with only one image, so they lack some information needed to figure out the distance of objects in the photo. The algorithm somehow figures that out by analyzing 'lines' in the photo to determine where objects meet the ground , and also using the horizon somehow as a baseline - although I did see on their website converted indoor photos where no horizon was visible. It also relies on 'training images' - probably images manually converted to 3D.
In any case, if you've seen those optical illusion illustrations where an object appears alternatively to be a solid brick or a missing brick like this one, sometimes even the brain can be fooled when translating a 2D image to 3D. So a computer algorithm would encounter that issue much more often, being that it lacks the power of interpretation based on prior knowledge that humans have.
That's why I think the algorithm cannot fully come up with a 3D model on its own. we'll have towait and see.
You can sign up for the fotowoosh beta on their website.
