MEAFNet: Multi-Expert Attention Fusion Network for Robust Grape Leaf Disease Classification

Abstract

Reliable detection of grape leaf diseases is critical for timely treatment and quality control in vineyard management. This paper presents MEAFNet, a novel Multi- Expert Attention Fusion Network designed to improve the reliability and accuracy of grape leaf disease classification. The proposed model integrates feature extraction mechanisms from the strengths of three pretrained convolutional neural networks such as MobileNetV2, EfficientNetB0, and DenseNet121 as parallel expert branches. A compact Transformer processes the three projected expert descriptors. Its self-attention operation allows each branch representation to be updated using evidence from the other two branches before classification. Experimental evaluations performed using a publicly available dataset of diseased grape leaves show that MEAFNet obtains an accuracy rate of 99.78% in classification within just 60 training epochs. The model achieves an F1-score of 1.000 on the Leaf Blight classes and maintains above 0.990 over all target classes. These results indicate the effectiveness and generalization capability of the proposed method in real-world agricultural environments. Overall, MEAFNet provides an effective framework for automated grape leaf disease classification and offers a promising foundation for intelligent disease monitoring in precision viticulture. 

Received: 18 August 2025 | Revised: 8 June 2026 | Accepted: 23 August 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work. 

Data Availability Statement

The data supporting the results of this research are openly available in the Grape Disease Dataset at https://www.kaggle.com/datasets/pushpalama/grape-disease. 

Author Contribution Statement

Hoang-Tu Vo: Conceptualization, Methodology, Writing – original draft, Supervision, Project administration. Nhon Nguyen Thien: Writing – review & editing, Visualization. Kheo Chau Mui: Software, Validation, Investigation. Phuc Pham Tien: Validation, Investigation, Writing – review & editing. Huan Lam Le: Formal analysis, Resources, Data curation.