COMPARATIVE ANALYSIS OF BINARY AND MULTICLASS POTATO LEAF DISEASE CLASSIFICATION USING VGG19 MODEL

Abstract

Agriculture in Nigeria, has been the source of livelihood yielding sustainable development across the country. However, potato farming in Nigeria faces numerous challenges such as unknown diseases and challenges in potato leaf disease classification. This study discovered a problem in potato leaf disease classification using VGG19 model in which binary class of potato leaf (potato early blight and potato late blight diseases) was not enough for dataset generalization. Therefore, this study aimed to conduct a comparative analysis of binary and multiclass potato leaf disease classification using VGG19 model. The research used comparative analysis tools to compare the result of the binary class (early blight and late blight leaves) and multiclass (early blight, late blight, virus disease and healthy potato leaves) in which VGG19 model with binary class obtained at epoch 40, training accuracy of 93.25% and loss of 0.2513, validation accuracy of 90.00% and loss of 0.2970 and testing accuracy of 91.67% and loss of 0.2735 and VGG19 model with multiclass obtained at epoch 40, training accuracy of 91.28% and loss of 0.3794, validation accuracy of 87.50% and loss of 0.3893 and testing accuracy of 91.67% and loss of 0.4956. The result showed that the higher the number of data classes in VGG19 model, the lower the training accuracy in VGG19 model. Finally, this work has achieved its aim and objective; and it can be evaluated for future study.