Computers may soon be able to tell what you’re doing based on images taken by your cellphone or wearable camera, MIT Technology Review reports. Using a database of 40,000 images taken by a single individual, a new piece of software has been able to identify—with 83% accuracy—what that person was doing at the time a photo was taken. The software categorizes images into 19 activity categories such as “driving” and “exercise.” It makes guesses based on visual cues, then incorporates factors such as location and time of day using a learning algorithm. Eventually, it could be used in apps that would learn a user’s habits and offer helpful suggestions on things like exercise and scheduling, the researchers told MIT Technology Review.