Researchers Develop an AI System to Train and Reward Dogs which respond to commands

Soon you will be able to use AI system to train your dog to stand or sit. Jason Stock and Tom Cavey, who are researchers from Colorado State University, have published a report on how to train artificial intelligence (AI) to reward dogs that will respond to certain commands.

The two computer science graduate students modeled image classification networks in a way that it could determine whether a dog was standing, sitting or lying down. The machine would then dispense a treat to a dog which will respond to the command through the adoption of the right posture. The treat will be served via a servo motor.

The students deployed Nvidia Jetson edge AI platform for real-time trick recognition and treats. They required dog images that displayed the three aforementioned postures. At least 20,000 images of varied sizes depicting the dogs in numerous positions were used in the research.

Nvidia told Venture Beat that: “It doesn’t yet work remotely; it is currently for in-person. But that would be an easy setup to make it a remote system. You might think of it as a system, or IP, to license for devices like Furbo. The researchers see many possible applications but haven’t committed to anything yet.”

The researchers used Nivdia’s Jetpack software development kit on Jetson for optimization. Nvidia TensorRT optimization libraries offered “noteworthy improvements in speed.”