Deep Learning for Plant Disease Detection: A Novel Convolutional Neural Network Approach Using the new dataset HealthySick Plants

¹ University Hassan II Casablanca, Faculty of Science Ben M’Sick, Laboratory of Artificial Intelligence and Systems (LIAS), Cdt Driss El Harti B.P. 7955, Ben M’Sick, Casablanca 20800, Morocco
² University Hassan II Casablanca, Faculty of Science Ben M’Sick, Laboratory Information Technologies and Modeling (TIM), Cdt Driss El Harti B.P. 7955, Ben M’Sick, Casablanca 20800, Morocco
³ University of Mons, Faculty of Engineering, ILIA, rue de Houdain 9, Mons 7000, Belgium

^* e-mail: fatimazahra.alaoui1-etu@etu.univh2c.ma

Abstract

This research explores a novel application of advanced deep learning techniques, with a particular focus on Convolutional Neural Networks (CNNs), to tackle the critical challenge of plant disease detection. A CNN-based classification model is proposed to accurately distinguish between healthy and diseased plants, utilizing state-of-the-art architectures. The model is trained and evaluated on a rigorously curated version of the Plant Village dataset, adapted in this study to focus on healthy versus sick plant classes. The dataset includes a diverse range of plant species and disease types, enhancing the robustness and generalizability of the model. The developed system achieved outstanding performance, with an accuracy of 99.53% and an AUC of 99.96%, demonstrating the reliability of AI-driven solutions in agricultural diagnostics. These findings highlight the transformative potential of integrating artificial intelligence into precision agriculture, contributing to early detection of plant diseases, reducing crop losses, and promoting sustainable and resource-efficient farming practices. This work represents a step forward in applying AI to support global food security and plant health management.