Bird Species Detection Using Deep Learning Techniques

Peacocks can significantly impact agricultural fields by damaging crops, necessitating an efficient detection system for timely intervention. This work introduces P-YOLOv5, a modified version of the YOLOv5 architecture designed for real-time object detection in resource-constrained environments. By replacing the original YOLOv5 backbone with a customized Mobile Net, which utilizes depthwise separable convolutions, P-YOLOv5 significantly reduces model size and computational complexity while preserving high detection accuracy. The model is evaluated on a detection dataset, demonstrating its effectiveness in tracking movement patterns related to environmental interactions. Experimental results show that P-YOLOv5 achieves a mean average precision (mAP) of 0.915 and a recall rate of 0.99, outperforming YOLOv5 in both inference speed and efficiency, with a 17% reduction in model size. These findings underscore P-YOLOv5 as a highly efficient solution for real-time object detection, offering a balance of speed, accuracy, and computational efficiency.