CLASSIFICATION OF PLANT LEAF DISEASES BASED ON CAPSULE NETWORK-SUPPORT VECTOR MACHINE MODEL

Machine learning methods adopted in classification of plant leaf diseases face challenges such as low accuracies, demand for huge training data among others. Some algorithms have been combined with other classification models to enhance performance but, scanty research has been done on the combination of capsule network (CapsNet) with other models to enhance classification performance. Tomato is a significant crop since it's used as a vegetable almost every household globally and its production is a source of livelihood to many people. In this study, a new model based on integration of CapsNet and support vector machine (CapsNet-SVM) classification model for classification of tomato leaf diseases was explored. Capsule network was optimized for feature extraction while the SVM model was used as a robust classifier. The main objective was to enhance the classification of support vector machine using automatic features extracted by the capsule network model. It was found that the CapsNet-SVM model was able to automatically extract features from the raw images and perform final classification. The proposed model was trained and evaluated using plantVillage dataset. Experimental results from the proposed model demonstrated a classification accuracy of 93.41%. The proposed showed significant performance compared to the existing methods used for classification of tomato leaf diseases.