SMS-based Dog Detection in Residential Area using YOLOv5 and ML.Net

The YOLOv5 object detection model is used in this paper to detect dogs, while the ML.Net is used to classify dogs in residential areas. A Raspberry Pi is used to transmit live video capture to be processed by the machine model algorithm. The method allows appropriate action by distinguishing between owned and stray dogs. The YOLOv5 model is trained on an extensive dog database to analyze real-time video footage taken by a camera. When dogs are detected outside their property, they are categorized, and SMS notifications are automatically issued to their owners. Furthermore, SMS alerts are sent to local authorities in response to the presence of stray dogs, assuring the protection of the animals and the community. This proactive approach tries to reduce the risks connected with roaming dogs while also encouraging ethical pet ownership. An optimum camera setup of 1.22 meters high with a camera angle of 24°, shows an 87.36% accuracy with a maximum detection distance of 3.05 meters. A bearable latency of 5 – 20 seconds is observed, considering both algorithms are running at the same time, the model had an average transmission time of 7.92 seconds for detection and SMS generation. 18.29 meters is found ideal outpost distance from the camera.