A CNN-SVM based model for detection and classification of Tomato leaf diseases

Abstract

to decreased crop yields and inferior fruit quality. Conventional disease detection techniques depend significantly on manual examination, resulting in delays and inaccuracies. This paper investigates the application of machine learning CNN-SVM methodology to create an automated system for the detection and classification of tomato leaf diseases. The research employs datasets from PlantVillage and Labeled_Features, consisting of more than 18,000 images of high-resolution tomato leaves afflicted by diverse diseases. The images underwent preprocessing, augmentation, and classification into three primary disease categories: Tomato Bacterial Spot, Tomato Leaf Curl Virus, and Tomato Mosaic Virus, in addition to healthy tomato leaves. The evaluation of our model attained the highest classification accuracy at 98.2%. The evaluation metrics, comprising precision, recall, and F1-score, averaged 98%, signifying the model's efficacy in differentiating between the leaves that have diseases and healthy leaves. Moreover, the study emphasizes critical attributes, including color variation, texture descriptors, and shape characteristics, which were essential in enhancing the model's performance. Notwithstanding the success, the study also delineates areas for enhancement, particularly in differentiating analogous diseases and mitigating environmental variability. The results highlight the efficacy of the CNN-SVM model in contemporary agricultural practices, providing efficient and economical solutions for real-time disease detection and management.