Data-Driven Insights into Potato Leaf Diseases: A CNN Approach to Early Blight Severity Classification

This abstract shall discuss the effectiveness of a designed model that could predict the severity of Early Blight disease in potato leaves using five Criticality Scales. Using an accurate estimation of the overall metrics regarding precision, recall, and F1-Score, the performance of the model is proven. For the Criticality Scale 1, it was set that precision, recall, and F1-Score were all around 98.59%. Accordingly, Criticality Scale 2 reaches 98.62%, and Scale 3 is marginally more potent with 98.65%. Performance is uniform across all scales: 98.63% for Scale 4 and 98.52% for Scale 5. The dataset contains 10,750 photographs, and the version carried out a usual accuracy of 98.60%. Further scrutiny using support and accuracy metrics shows the model developed on altogether images of 10,730 with support values at 2,130 for Scale 1, 2,180 for Scale 2, 2,220 for Scale 3, 2,190 for Scale 4, and 2,030 for Scale 5. The impressive thing is that it achieves an accuracy of 99% for all the scales, which shows that the model is reliable. Such results demonstrate the ability of the model to help farming professionals identify and monitor the disease, which leads farmers to strengthen their agricultural practices and thus leads to the achievement of improved agricultural outcomes. High values of precision, recall, and accuracy indicate that the model has great potential as a useful tool in agricultural disease management.