Make-3D is a service out of Stanford that allows you to upload any photo and it converts it into a 3D movie (or more generally, a 3D model that you can explore if you have the appropriate browser plugin). I uploaded this photo:
and a few minutes later, I was the proud owner of this:
This is how it works:
Our software uses a breakthrough technology in machine learning. It estimates depths from the single image by using our monocular vision algorithm, developed in 2005. It captures a variety of monocular cues and learns the relations between different parts of the image using a machine learning technique called Markov Random Field (MRF). Our algorithm first divides the image into small patches and analyzes them at multiple scales to estimate each of the patches’ 3-d location and 3-d orientation. More details could be found here.
The smarter ones amongst you would have realized that what the above means is that the program is pretty much guessing. And it can make mistakes. For example, I uploaded this photo:
and ended up with this 3-D model:
As you can see, the algorithm seems to have put a huge hole in the tower, and Ronak appears to be doing more of a Spiderman routine than regular rappelling.
By the way, the site is very easy to use. Try it yourself.
