What makes a nose a nose?
In the field of computer vision, scientists frequently find themselves asking such questions. Creating an algorithm to help computers identify shapes has turned out to be a task ridden with problems.
With traditional shape-identification methods, an algorithm would have to cover thousands of nose-shape variations, not to mention exhaustive definitions for the numerous other objects a computer may have to identify. A human can identify a nose as a nose regardless of deformities, lighting conditions, etc., but it’s proven infinitely more difficult for a computer to achieve this same flexibility.
However, a new computer-vision technique has taken a huge step to overcoming this barrier. Read the full article about the closest we’ve come to giving computers sight.