Smartwatches and dedicated wrist-based activity trackers have become popular over the past few years. They’re decent at counting steps during walking and running, but they’re not very good at tracking most athletic activities. At Carnegie Mellon University researchers have developed a camera-based system that simply watches people at the gym doing exercises and automatically classifies their activities and counts the repetitions.
The computer vision software recognizes repetitive motions and assumes those are exercises being performed over and over. Different exercises look different and with a bit of training, the software was taught to differentiate between many types.
Only one camera is enough to monitor a bunch of people working out at a gym, but the system uses low-resolution video to do its job and so can’t be used to recognize individuals, a nice privacy benefit.
While it has been applied in a crowded gym at Carnegie Mellon, the software can also run on smartphones and other devices, and so may soon be available as an app, as part of exercise equipment, and at your local gym.
Here’s a video that shows off the new software:
Study in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies: GymCam: Detecting, Recognizing and Tracking Simultaneous Exercises in Unconstrained Scenes
Technology info page: GymCam…