Performance of Fused Multi-Stage Transfer Learning in Classification of Grape Diseases
Isaac Angelo Dioses, Reonel Ferreria, Michael John P. Robles, Fitzerald D. Lim, Mark Joseph Asuncion, Jolan Sy
Abstract
Grapevine diseases significantly impact the global wine and fruit industry, necessitating efficient and accurate early detection methods. This study proposes a multi-staged transfer learning (MSTL) approach for classifying four grape leaf conditions: Leaf Blight, Black Measles, Black Rot, and Healthy leaves. Using a custom deep learning pipeline that integrates image preprocessing, convolutional layers, pooling, dropout, and fine-tuned dense layers, the model leverages knowledge from pre-trained networks to enhance performance on limited agricultural datasets. The proposed method achieved an overall accuracy of 98%, with macro-averaged precision, recall, and F1-scores all at 0.98. Results from the confusion matrix and classification report confirm the model’s reliability and robustness across all classes. The findings demonstrate the effectiveness of MSTL in agricultural disease classification and suggest its potential for real-world deployment in smart farming systems and mobile applications.