Multi-Module Deep Learning and IoT-Based Pest Detection System Using Sound Analytics in Large Agricultural Field

Agriculture has progressed to an advanced stage worldwide and into an economical source of sophisticated agricultural practices. It is a significant milestone in the history of agriculture. The broad embrace of information and today's advanced technology has paved the way for this evolution. But pests cause substantial harm to farming, which has serious financial, ecological, and societal repercussions. That emphasizes the significance of computerized systems for early pest detection. Agricultural issues have recently emerged as a significant focus for machine learning-based research. This study offers an efficient Internet of Things (IoT)-enabled pest detection system that uses pest sound analytics in large agricultural fields. The system is built on a model called a Graph-based Pyramid Attention-multi module deep convolutional neural network (GPA-Net). The GPA-Net platform comprises a variety of essential subcomponents, a couple of which are the CSPNet, CTA, and SPA. The GPA-Net model has been trained using the sounds of the pests. The outcomes of the demonstrations revealed that the recommended GPA-Net model successfully identified the pest and outscored the state-of-the-art models YOLOv5, Faster RCNN, and ResNet-50 with 99.75% accuracy and an F1 score of 98.99%.