Development of CNN-Based Model for Mango Leaf Disease Classification

Food security has become an urgent topic that needs more focus. It is one of 17 sustainable development goals. Specifically, SDG 2: zero hunger. We need effective solutions from multidisciplinary domains to achieve this goal. The advancement and growth of artificial intelligence could also provide a promising solution to achieving zero hunger. This paper proposes a solution to promote sustainable agriculture practices. We chose the classification of mango leaf disease as our case study. We successfully develop an accurate and efficient convolutional neural network-based model as a solution. The proposed model achieved 99.5% accuracy and outperformed the majority state-of-the-art models, only losing to DeiT, which achieved 99.75% accuracy. However, the proposed model comes with 739,080 parameters, which are 4.73 and 115.95 times smaller than MobileNetV2 and DeiT, respectively. We believe this finding could assist farmers in accurately classifying mango leaf disease, thereby enabling timely and precise preventive action. This leads to a decrease in material loss and the preservation of plant productivity. Leveraging this achievement, forthcoming initiatives seek to disseminate these technologies across multiple plant diseases.