A Comprehensive Survey on Real-Time Animal (Dog) Detection System Using Artificial Intelligence Methods

Stray dogs pose public health and safety risks in communities worldwide. Humane management of stray dog populations requires accurate monitoring. This paper reviews research on applying computer vision and deep learning to detect and count stray dogs from CCTV footage. A comprehensive literature search identified 32 relevant studies from the past 10 years. Current techniques utilize convolutional neural networks (CNNs) like YOLO and Faster R-CNN for stray dog detection. Studies report precision of 80–95% and recall of 70–80% on small customized datasets. However, performance drops significantly when tested in diverse real-world environments. The main challenges include distinguishing dogs from other objects and handling varying backgrounds, weather conditions, and lighting. Proposed future work focuses on expanding datasets, testing generalizability across locations, incorporating sensors to enhance capabilities, and deploying optimized models on embedded systems for real-time use. Overall research progress is promising but heterogeneous datasets, model optimization, and rigorous validation in uncontrolled settings remain open problems. Advanced deep learning and domain adaptation techniques tailored to stray dogs in unconstrained environments are needed to translate high accuracies to field deployment.