Tomato Disease Classification using Different Deep Learning Approaches

Abstract

Accurate identification of tomato diseases from leaf images is a complex task that presents significant challenges even for agricultural experts. This study addresses this issue by developing a deep learning-based classification system using Convolutional Neural Networks (CNNs). Specifically, we investigated the performance of various MobileNet architectures (V2, V3 Small, and V3 Large) trained from scratch on a standardized dataset of 224x224 pixel images. The models were evaluated based on their ability to classify leaf images into distinct disease categories. Experimental results demonstrated that MobileNetV3 Large achieved the highest test accuracy of 96.9%, outperforming MobileNetV3 Small (96.34%) and MobileNetV2 (93.8%). Through hyper parameter tuning and comparative evaluation, the MobileNetV3 Large model was selected as the optimal classifier for deployment. The findings suggest that efficient Mobile Net architectures provide a robust solution and light weight model for automated plant disease detection, offering a viable tool for precision agriculture.