Deep learning-based potato leaf disease classification and severity assessment

A deep learning model was developed to detect, classify, and assess the severity of potato leaf diseases using 4200 images captured with a smartphone. The dataset, including healthy and diseased leaves affected by early and late blight, was preprocessed with techniques like median filtering and data augmentation. For disease identification, three classes, early blight, healthy, and late blight, were targeted. For severity level detection, potato leaf images with early and late blight were categorized into six classes: EB-low, EB-moderate, EB-severe, LB-low, LB-moderate, and LB-severe. The proposed model's performance was compared with AlexNet and VGG16. The model, using a CNN architecture with 1 fully connected and 11 convolutional layers, achieved an overall accuracy of 99% for disease classification. For severity detection, the model reached a 96% average accuracy. Performance was compared to AlexNet and VGG16, with the proposed model outperforming both in all metrics. The softmax activation was used in the output layer to classify potato leaf diseases as well as categorize the severity level of diseased leaf images.