A Hybrid Model using MobileNetv2 and SVM for Enhanced Classification and Prediction of Tomato Leaf Diseases

Abstract

This study proposes an innovative method for automated categorising tomato leaf diseases using a hybrid model that combines deep learning and machine learning approaches. The proposed model integrates Convolutional Neural Networks (CNNs) pre-trained model (MobileNetv2) for feature mining and SVM for disease multi-class classification and prediction. A comprehensive dataset of 10,000 tomato leaf images was collected, preprocessed, and utilized for model training and evaluation. Experimental results demonstrate the efficacy of the hybrid model, achieving an impressive accuracy ranging from 88% to 99% in multi-class disease image classification. According to the comparative analysis, the relevant models' accuracies range from 70% to 97%. The outcome further endorses the proposed techniques' superiority. It highlights the necessity of choosing the appropriate model architecture and optimization strategies for obtaining high accuracy in image classification tasks, indicating its potential for real-world deployment. This study adds to the advancement of agronomic know-how by providing an efficient and accurate tool for tomato disease management, enabling early detection, and supporting sustainable crop production.