Thanks to the efforts of a small band of researchers, it may soon be possible to find an image on the Internet without relying on a human description of it. Two computer scientists from Google are among the pioneers seeking to advance the massive image database created by Stanford University's Fei-Fei Li using "neural network" algorithms designed to mimic human brain functions. Andrew Ng and Jeff Dean say their visual recognition technology is already making it possible for the ImageNet database -- which holds more than 14 million images -- to categorize unlabeled objects with surprising accuracy.

Related Summaries