Title: Deep Transfer Learning for Accurate Classification of Grape Leaf Diseases
Authors: Malak S. Abu Mettleq and Samy S. Abu-Naser
Volume: 10
Issue: 6
Pages: 1-9
Publication Date: 2026/06/28
Abstract:
Artificial intelligence (AI)-based plant disease detection has become an essential component of modern precision agriculture, contributing to improved crop management, enhanced productivity, and sustainable food security. This study presents an advanced deep learning framework for the automated classification of grape leaf diseases using a fine-tuned Xception convolutional neural network. A balanced dataset comprising 8,000 grape leaf images representing four classes-Black Rot, ESCA, Leaf Blight, and Healthy leaves-was employed to ensure robust and unbiased model training. The dataset was divided into training (60%), validation (20%), and testing (20%) subsets. To enhance classification performance and generalization capability, the proposed framework incorporates transfer learning, extensive data augmentation, batch normalization, dropout regularization, and combined L1-L2 weight penalties. Model effectiveness was assessed using multiple evaluation metrics, including accuracy, precision, recall, and F1-score, together with confusion matrix analysis and Receiver Operating Characteristic (ROC) curves for comprehensive performance validation. Experimental results demonstrate exceptional classification performance, achieving a test accuracy of 99.90% and an F1-score of 99.85%. While the obtained results indicate strong predictive capability, additional analyses emphasize the importance of evaluating the model under more diverse environmental conditions to address potential challenges associated with real-world deployment. Comparative evaluation against existing approaches further confirms the superiority and reliability of the proposed framework. The findings highlight the potential of the developed model for integration into precision agriculture applications, including mobile and edge-computing platforms, enabling efficient real-time monitoring and early diagnosis of grape leaf diseases.