Biotic Stress and Severity Classification of Pear Leaves Disease with Ensemble Learning

Leaf disease frequently arises as pears plants mature. With a particular emphasis on categorizing the pear leaf disease, the proposed research work suggests the implementation of the CNN architecture that we use in this study to create this disease identification and classification method comprised of ResNet50, InceptionV3, VGG16, and VGG19. These neural networks have been pre-trained on the ImageNet dataset. The ensemble methodology that we describe produces final predictions for the test samples based on the recall, precision, accuracy, and f1-score scores of four standard evaluation metrics. Even though there is an increase in the amount of accuracy when the models are ensembled, the remarkable combination is InceptionV3 with VGG16 with an accuracy of 77.25%, precision of 79.34%, recall of 77.25% and F1 score of 76.50% and with the severity classification as accuracy 45.29%, precision 46.88%, recall 45.29% and F1 score 45.50% respectively.