Comparative Analysis of SVM and CNN Techniques for Tomato Leaf Disease Classification

Abstract

Tomatoes are among the most widely cultivated crops worldwide and are valued for their taste, nutritional benefits, and culinary versatility. However, tomato plants are sensitive to various leaf diseases that can significantly reduce yield and quality, making early and accurate disease identification essential for effective management. This study involves two machine learning techniques to categorize tomato leaf diseases using a dataset of 10,000 labelled images representing ten categories, including healthy leaves and bacterial, viral, and fungal infections. A Convolutional Neural Network (CNN) model was developed in Python using TensorFlow with a pretrained DenseNet121 architecture, enabling automatic feature extraction and classification. In contrast, a Support Vector Machine (SVM) model relies on manually extracted features such as color, texture, and shape. Model performance is evaluated using accuracy, precision, recall, and F1-score metrics. The CNN outperformed the SVM, achieving an accuracy of 93.35% compared to 78% for the SVM. These results highlight the superiority of deep learning approaches for plant disease detection. The proposed CNN-based system can assist farmers in timely disease diagnosis, reduce crop losses, and minimize pesticide usage. Furthermore, the model can be deployed in real-world agricultural settings through integration with mobile-based platforms or Internet of Things systems for on-site disease monitoring and decision support.