Researchers Visualize Image Recognition Output to See Things From a Computer’s Perspective
Even the best object-recognition systems, however, succeed only around 30 or 40 percent of the time — and their failures can be totally mystifying. Researchers are divided in their explanations: are the learning algorithms themselves to blame? Or are they being applied to the wrong types of features? Or — the “big-data” explanation — do the systems just need more training data?
To attempt to answer these and related questions, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory have created a system that, in effect, allows humans to see the world the way an object-recognition system does. The system takes an ordinary image, translates it into the mathematical representation used by an object-recognition system and then, using inventive new algorithms, translates it back into a conventional image.
It’s not quite Terminator, but it’s not that far off either: Check out Atlas, a new, 6-foot, 2-inch-tall humanoid robot designed for a contest being held by US Defense Department. The 290-pounds machine is being called “one of the most advanced humanoid robots ever built,” in no small part due to its 28 different hydraulic joints and freakishly good balance.