MobileNet-Based Tomato Leaf Disease Detection: Enhancing Agricultural Diagnostics through Deep Learning

Tomato leaf diseases pose a serious threat to tomato yield and quality, which is paramount to supporting sustainable agriculture worldwide with timely disease detection. Tomatoes remain one of the most eaten and cultivated vegetables in the world, with global production reaching roughly 186 million metric tonnes every year. The most important production sites are China, the United States and the Mediterranean countries which account for a large part of world supply. Tomatoes remain a cornerstone of diets across the globe, as evidenced by worldwide processing volumes that reached nearly 44.2 million metric tonnes (MMT) in 2023 despite recent hurdles brought on by climate variability and rising production costs. A deep learning model based on MobileNetV2 architecture, optimised by Adam optimizer for tomato leaf disease identification and classification is proposed in this study. The model was trained on a rich dataset of ten classes (nine disease categories and healthy leaves) and exhibited a high training accuracy of 99.92% with a validation accuracy of 98.40%, thus enabling it for real-life practical applications on mobile devices. A comparison of simulation using alternative optimizers, namely SGD, Nadam and RMSProp was made that shows Adam achieved a lower error rate in less time thus outperforming other optimizers which makes it preferable for complex real-world agricultural conditions. Our results highlight MobileNetV2 as a simple efficient tool for in-field diagnostics to aid in early disease detection and loss prevention. The lightweight, efficient design of the model can be scaled and adapted for any other crop with little retraining, demonstrating promising capabilities to be transferred to other horticultural diseases beyond tomato leaf ones. This research demonstrates the increasing role of AI-driven agriculture in realizing food security through accurate, scalable diagnostics.