And when asked to distinguish objects at a higher level of detail, Sketch-a-net really set itself apart from humans, correctly differentiating among seagulls, flying birds, standing birds, and pigeons 42.5 percent of the time. Humans, on the other hand, were only correct 24.8 percent of the time. And as this technology becomes more and more robust, researchers say that it could be combined with existing touchscreen capacities to create more precise search results — after all, sometimes, showing really is better than telling (or in this case, drawing is better than typing).
According to researchers, part of the computer program’s success comes from its ability to identify the order in which the strokes in a drawing were made. As a result, if you’re a particularly erratic sketcher (or an abstract artist), the computer may meet with less success. Researchers also note that their use of “a network architecture designed for sketch rather than natural photo statistics” helped them achieve great success in identification.
Study co-author Dr. Timothy Hospedales told GizMag, “It’s exciting that our computer program can solve the task even better than humans can. Sketches are an interesting area to study because they have been used since prehistoric times for communication and now, with the increase in use of touchscreens, they are becoming a much more common communication tool again.”
So the next time you play Pictionary, consider bringing a computer into the fold. It may wind up beating all your human friends.
