An Automated CNN Framework for Multi-Class Apple Leaf Disease

Abstract

Apple is one of the most economically valuable fruit crops cultivated worldwide; however, its productivity is significantly affected by leaf diseases such as apple scab and cedar apple rust. Conventional disease identification methods are often labor-intensive, time-consuming, and highly dependent on human expertise, resulting in subjective diagnoses. To address these limitations, this study proposes an automated apple leaf disease classification system based on a Convolutional Neural Network (CNN). A subset of the PlantVillage dataset comprising 1,200 images was utilized, including 400 healthy apple leaves, 400 apple scab images, and 400 cedar apple rust images. The dataset was divided into 80% training and 20% testing subsets. Image preprocessing consisted of resizing, pixel normalization, brightness adjustment for contrast enhancement, and data augmentation to improve model generalization. The CNN model was trained using the Adam optimizer with the categorical cross-entropy loss function. Experimental results demonstrated an overall classification accuracy of 98% without evidence of significant overfitting. The classification report showed that the healthy class achieved 0.97 precision, 0.97 recall, and 0.97 F1-score; the cedar apple rust class achieved 0.99 precision, 1.00 recall, and 0.99 F1-score; while the apple scab class obtained 0.97 precision, 0.96 recall, and 0.97 F1-score. Most classification errors occurred in the apple scab class due to its visual similarity to healthy leaves. These findings demonstrate that the proposed CNN architecture effectively learns discriminative visual features for multiclass apple leaf disease classification. The proposed system has strong potential as an early disease detection tool to support precision agriculture and sustainable apple crop management. Future work will focus on employing deeper CNN architectures, transfer learning techniques, and image segmentation methods to further improve the recognition performance of challenging disease classes, particularly apple scab.