AI-Driven Detection of Tomato Leaf Diseases for Sustainable Agriculture

Abstract

This study explores a novel approach for detecting diseases in tomato leaves through the application of neural networks, aiming to enhance early diagnosis and management strategies for farmers and plant pathologists. The research investigates nine prevalent diseases affecting tomato foliage, including Early Blight, Late Blight, Septoria Leaf Spot, Target Spot, Yellow Leaf Curl Virus, Bacterial Spot, Spider Mites, Leaf Mold, Tomato Mosaic Virus, and Healthy leaves, using pre-trained deep learning models, ResNet-34 and VGG16. A diverse dataset of tomato leaf images, exhibiting various disease symptoms under field and curated conditions, was pre-processed, labeled, and split into training (80%) and testing (20%) sets to fine-tune the models. Evaluation of the testing dataset revealed that ResNet-34 achieved a higher accuracy of 99% compared to VGG16’s 89%, demonstrating superior performance in disease classification. Precision, recall, and F1 scores further confirmed ResNet-34’s robustness, averaging 0.99 across classes. These findings highlight the efficacy of deep learning in agricultural disease detection, contributing to sustainable practices by enabling timely interventions, reducing crop losses, and minimizing pesticide use. The study underscores the potential of AI-driven solutions to transform tomato cultivation, paving the way for scalable, real-time applications in resource-constrained farming environments.