Researchers built an AI that recognizes and rewards good doggos
The team trained NVIDIA’s Jetson edge AI using more than 20,000 images of our furry friends.
You might soon be able to use an AI system to help train your dog to sit. A pair of researchers from Colorado State University have developed an artificial intelligence system that detects when a dog is sitting, standing or lying. If your furry friend takes up the right position upon your command, the system will reward your pooch by automatically dispensing a treat via a servo motor.
Jason Stock and Tom Cavey, who are computer science grad students, used NVIDIA’s Jetson edge AI to create the system. They trained it using more than 20,000 labeled images of dogs in various positions.
The duo’s willing volunteer was Henry, Cavey’s Australian Shepherd. Using a camera, the system was able to analyze Henry’s position in a split second. It had up to 92 percent test accuracy at around 39 frames per second, according to Stock and Cavey’s paper.
The researchers suggest that dog trainers and owners could use the system as a teaching aid, so you might be able to train your very good dog remotely when you eventually get back to the office. It may also give you another way to treat your pupper while you’re away from home.