Stray Dog Detection System using YOLOv5

Abstract

Stray dogs present significant public health and safety risks, particularly in developing countries like India, where the stray dog population is the largest globally. This paper details the implementation of a Stray Dog Detection System using the YOLOv5 object detection model to automatically detect and track stray dogs in real time via CCTV feeds. YOLOv5’s high accuracy and real-time processing capabilities make it well-suited for detecting stray dogs in complex, crowded urban environments. The system leverages a YOLOv5 model trained on custom datasets tailored to local conditions, including specific dog breeds and deployment environments. It integrates an alert mechanism that triggers when stray dog populations surpass predefined thresholds, allowing timely interventions. Additionally, the system incorporates geographic mapping to provide data-driven insights for municipal authorities to manage stray populations effectively and ethically. Experimental results demonstrate an F1 score of 0.97, validating the system’s robustness for practical deployment. This paper discusses system architecture, implementation, and performance, highlighting its scalability and cost-effectiveness for humane stray dog population control.