AI may be able to train dogs in the form of snacks if you follow instructions such as sitting or lying down. Colorado State University graduate students developed a system that instantly detects dog posture and provides snacks according to instructions.
Colorado State University graduate students have developed a system that can determine the three states of a dog sitting, standing, or prone using the NVIDIA Jetson AI platform. It is said that more than 20,000 photos of dogs in various positions were used for AI learning.
When it detects that the dog is in the correct posture, the servo motor is activated and snacks are served. In the test using dogs, it is said that it was able to infer at 39 frames per second and detect the posture with 92% accuracy.
Of course, this research is not for the development of snack dispensers, but an AI efficient learning and optimization technique. In order to reinforce the dog’s behavior, quick reasoning is required, and the use of Raspberry Pi was also reviewed, but Jetson said it was the most meaningful.
In addition, there is a possibility that a product using this structure will appear, but they plan to conduct further research to detect the shape of dog ears and tail. If this happens, I am hoping that I can read the feelings of the dog. Related information can be found here .