Comparative Analysis of Deep Learning Models for Tomato Leaf Disease Classification

The present work compares the performance of four Convolutional Neural Networks (CNNs), namely AlexNet, VGG16, MobileNetV2, and GoogleNet, for the classification of tomato leaf disease on the Plant Village dataset, consisting of 10,000 training and 1,000 testing images in 10 classes. Numerous preprocessing steps, such as resizing, normalization, and data augmentation, were performed. Among these, the validation accuracy was the highest (99.65 %) for AlexNet but was overfit and therefore not best suited for deployment. VGG16, with the lowest accuracy (91.55 %), performed poorly because of its high-complexity architecture and slow convergence. GoogleNet (94.10 %) and MobileNetV2 (94.80 %) demonstrated a balanced trade-off between accuracy and computational efficiency, with MobileNetV2 emerging as the most promising candidate for real-time implementation due to its lightweight architecture and fast inference capabilities. The work is an indication that there is a need to customize CNN architectures to diagnose farm diseases while taking deployment challenges like hardware and real-world variability into account.