I am a proud CMU alumnus. When I was there, I always had the sense that I was surrounded by incredibly smart people. Engineers, mathematicians, economists, heck, even the drama students do great things (especially my bud Adam Donshik).
I was browsing YouTube’s popular videos today, and in between clips from Saturday Night Live, the Daily Show, and Mentos funniness, I noticed this amazing video of a system generating 3D images from 2D photographs. I must say, it’s much cooler than being responsible for the emoticon!
It is a interesting idea that was presented at CVPR this year, but it isn’t 3D reconstruction. What they don’t tell you in the video is that they require a training set of images, and they are looking for “sky”, “ground” and “vertical” based on color, texture and vanishing points. The false alarm rate is high, and it isn’t applicable to all images. Cool video, though.