Brown University researchers have developed software that can identify simple sketches in real time, which they note is the first program that enables semantic understanding of abstract sketches.
Brown professor James Hays says the key to making the program work is a large database of sketches, which can be used to teach a computer how humans sketch objects. The researchers developed the database by devising a list of 250 categories of everyday objects that people might be inclined to develop. They then used Mechanical Turk to hire people to sketch objects from each category, and accumulated 20,000 sketches. The data was fed into existing recognition and machine-learning algorithms to teach the program which sketches belong to which categories. The researchers then developed an interface in which users input new sketches and the computer tries to identify them in real time.
The program currently identifies sketches with about 56 percent accuracy, compared to 73 percent for humans. "The gap between human and computational performance is not so big, not as big certainly as it is in other computer-vision problems," Hays says.
From Brown University
View Full Article
No entries found