EffResViT-SE FusionNet: A Hybrid Deep Learning Framework for Accurate Classification of Coffee Leaf Diseases

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology
• Artificial Intelligence

Background:

• Coffee cultivation faces significant threats from leaf diseases like Leaf Rust, Phoma, Cercospora, and Leaf Miner.
• Existing diagnostic methods (visual inspection, PCR, ELISA) are often slow, expensive, and require expert knowledge, hindering widespread application.
• These diseases severely impact coffee plant health, reducing photosynthetic efficiency, causing defoliation, and diminishing crop yield and quality.

Purpose of the Study:

• To develop a novel, automated system for the early and accurate detection and classification of coffee leaf diseases.
• To address the limitations of traditional diagnostic techniques by leveraging advanced deep learning methodologies.
• To provide a scalable and precise solution for disease management in coffee farming.

Main Methods:

• Proposed EffResViT-SE FusionNet, a hybrid deep learning framework combining EfficientNetB3 and ResNet50 with Squeeze-and-Excitation (SE) blocks and a Vision Transformer (ViT).
• Trained the model on a large dataset of 58,555 coffee leaf images across five categories: Healthy, Miner, Leaf Rust, Cercospora, and Phoma.
• Utilized Adam optimizer, a learning rate of 0.001, batch size of 32, and 80 training epochs for model convergence.

Main Results:

• Achieved an outstanding overall classification accuracy of 99% for coffee leaf disease detection.
• Demonstrated high performance across all classes with precision, recall, and F1-scores ranging from 98% to 99%.
• Outperformed baseline models (ResNet50, EfficientNetB3, ViT) significantly, with accuracy improvements of up to 5%.

Conclusions:

• EffResViT-SE FusionNet offers a powerful, precise, and scalable solution for identifying coffee leaf diseases.
• The hybrid architecture effectively integrates local and global feature extraction for superior diagnostic accuracy.
• This AI-driven approach supports timely disease management, contributing to sustainable coffee agriculture and farmer livelihoods.