Automatic Detection and Classification of Wild Animal Species Using YOLO Models

Using a camera trapping framework, which employs cameras activated by sensors to record a burst of pictures of animals in their habitat, it is feasible to monitor animals in the open. This paper explores the effectiveness of YOLOv3 and YOLOv5 object detection algorithms in automated animal tracking through the use of computer vision. A dataset of six animal classes was utilized to train and test both models. The results demonstrate that YOLOv5-m achieves an 81% detection accuracy rate, surpassing that of YOLOv3. In addition, the two models were evaluated based on detection speed and model size, with YOLOv5-m providing higher accuracy and comparable speed and model size to YOLOv3. Overall, these findings highlight the potential of computer vision and deep learning in automated animal tracking, and emphasize the importance of selecting appropriate algorithms for various use cases.