Deep Learning-Based Classification of Potato Leaf Diseases for Sustainable Agriculture

Agriculture is one of the most vital sectors that determine the survival of man on earth, and potatoes are a major crop grown in Bangladesh. But the increasing incidence of diseases, including early blight, late blight, and hollow heart, is a major threat to potato production that costs farmers and harms their means of living. In this work, we tackle this problem by developing an efficient automated tool for disease detection with the help of advanced machine learning techniques. In this work, we have performed a comparative analysis of CNN models using a dataset of 6,666 potato leaf images for disease classification. The models consisted of a basic CNN and other pre-architectures like VGG16, Weighted, Xception, DenseNet121, and MobileNetV2. It was observed that high classification accuracies were shown by 2D CNN as well as VGG16 and were shown to be able to achieve 99.7% and 99.1%, respectively. The latter will be used to diagnose plant diseases and possibly increase agricultural productivity.