Four-Layered IoT-WSN Framework for Data Acquisition and Pest Classification Using Efficientnet in Sustainable Smallholder Farms

Smallholder agriculture, a critical source of livelihood for a significant portion of the global population, faces numerous challenges in adopting modern technology. In developing countries, nearly 86% of farmers rely on small plots of land, contributing to low yields due to traditional farming methods. To address these challenges, this paper proposes an architecture that connects Internet of Things (IoT) devices and Wireless Sensor Networks (WSN) for reliable data acquisition, alongside a deep learning model for efficient pest classification, aimed at improving the productivity and sustainability of smallholder farms. The proposed solution integrates IoT and WSN to create an efficient system for collecting real-time environmental parameters, such as soil moisture, temperature, humidity, and pest activity, using sensors deployed across farmlands. A comprehensive dataset is then used to train a deep learning model for predictive pest classification. The model, based on EfficientNetB4, achieved an impressive accuracy of 90.17%, with precision, recall, and F1-score values of 90.10%, 90.17%, and 90.02%, respectively, demonstrating the efficacy of deep learning in pest detection. This paper highlights the role of deep learning in enhancing pest classification and presents a sustainable architecture for data acquisition, enabling farmers to gain valuable insights into pest activity and optimize their practices. This approach not only contributes to the "United Nations’ Sustainable Development Goal Number 2: Zero Hunger", promoting sustainable agriculture, but also significantly improves food security and farming resilience, revolutionizing smallholder agriculture to meet global food production demands.