Like most smartphone users, you probably have hundreds of photos stored on your phone. But what if you’re looking for something specific, perhaps that selfie you took a few years ago with the beach in the background? Instead of sorting through all those photos, wouldn’t it be neat if you could just type in the word "beach" and just pull up photos with beaches in them?
Soon, you’ll be able to, thanks to researchers at Purdue University who are using artificial intelligence to make your phone recognize objects in pictures you’ve taken, or even in your current environment.
Purdue calls this idea deep learning because it requires a machine to basically do what the human brain does when it processes information, recognizes something, and spits out the name for it.
Deep learning is already being used elsewhere on the Internet, with tags being automatically applied to objects and faces in photos, but it’s never been possible before on mobile devices because it requires a heavy amount of computer processing.
However, the Purdue team developed new software and hardware and showed how a smartphone could run deep learning on its own. The idea is that the software’s algorithm looks at an image and separates it into multiple layers.
For example, for recognizing faces, each facial feature (such as eyes, nose, mouth) is given its own layer, and the algorithm studies each layer until the whole can be identified. After identification, the program tags the objects and faces in the photo with the appropriate name.