Home // EMERGING 2025, The Seventeenth International Conference on Emerging Networks and Systems Intelligence // View article

Intelligent Pest Identification for Precision Agriculture using Deep Learning

Authors:
Anika Bhat
Atul Dubey

Keywords: Wheat Pest Identification; Deep Learning; Machine Learning; MobileNetV2; ConvNeXtLarge; Internet of Things.

Abstract:
Global wheat production suffers annual losses of ~157 million metric tons due to pests, causing food insecurity and economic damages exceeding $70 billion. Traditional detection methods, such as manual inspections, are slow, labor-intensive, and often fail to identify early infestations, forcing farmers to use excessive pesticides. To address this, an image analysis system driven by Artificial Intelligence (AI) was developed, trained on pest imagery, and deployed via an accessible web application, enabling early detection to prevent crop losses. The IP102 dataset with wheat pest categories only was used to train the Machine Learning (ML) models. Two approaches were used to build an ML model that can detect wheat pests. The first employed transfer learning on MobileNetV2, and it gave the best validation accuracy of 55.32%. The second used ConvNeXtLarge to extract robust image features of 9 categories of wheat pests. Four ML algorithms, K-Nearest Neighbors (KNN), Random Forest, Multi-Layer Perceptron (MLP), and XGBoost, were trained and evaluated. The MLP model, optimized with 30 epochs and a learning rate of 0.001, achieved the highest validation accuracy of ~79% and test accuracy of ~75%. The system was integrated into a user-friendly web application, paired with a low-cost, WiFi-enabled camera device for field image capture. This system facilitated early-stage pest detection, enabling farmers to remotely monitor and take preventative measures promptly. This AI-driven model can contribute to efficient, sustainable, and precise agricultural practices and bolster global food security.

Pages: 6 to 12

Copyright: Copyright (c) IARIA, 2025

Publication date: September 28, 2025

Published in: conference

ISSN: 2326-9383

ISBN: 978-1-68558-292-0

Location: Lisbon, Portugal

Dates: from September 28, 2025 to October 2, 2025