EDISP: a hybrid CNN-ViT framework for robust maize leaf disease detection and classification

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Maize foliar diseases cause significant yield losses, necessitating timely and accurate identification.
• Manual disease inspection is subjective and less effective.
• Deep learning models like CNNs and ViTs show promise for automated diagnosis but have limitations.

Purpose of the Study:

• To develop a hybrid deep learning framework, EDISP, integrating CNNs and ViTs for enhanced maize foliar disease detection.
• To overcome the limitations of standalone CNNs (local focus) and ViTs (large datasets, missed fine details).

Main Methods:

• Developed EDISP, a hybrid framework combining CNNs for local feature extraction and ViTs for global contextual learning.
• Trained EDISP on a diverse dataset of controlled and real-field maize leaf images with thorough preprocessing.
• Utilized stratified splitting for training, validation, and testing to ensure generalization.

Main Results:

• EDISP achieved superior performance over standalone CNN and ViT models.
• Achieved high classification accuracy (99.40%), precision (99.43%), recall (99.38%), and F1-score (99.40%).
• Successfully identified various maize diseases with minimal errors and demonstrated robustness on independent datasets.

Conclusions:

• The EDISP hybrid architecture offers a robust and accurate solution for automated maize leaf disease detection.
• Demonstrates the potential of hybrid deep learning in precision agriculture for scalable disease management.
• Provides a valuable tool for farmers lacking expert diagnostic resources.