Deep Learning-Based Barley Disease Quantification for Sustainable Crop Production.

Area of Science:

• Plant Pathology
• Computational Biology
• Agricultural Science

Background:

• Net blotch, caused by Drechslera teres, significantly impacts barley (Hordeum vulgare) yield.
• Accurate quantification of disease severity is crucial for crop management and research.

Purpose of the Study:

• To develop and validate a deep learning model for quantifying net blotch disease symptoms in barley.
• To assess the model's performance using Cascade R-CNN and U-Net architectures.

Main Methods:

• Utilized a dataset of annotated barley leaf images for training.
• Employed Cascade R-CNN and U-Net deep learning architectures for symptom detection and quantification.
• Validated the model against traditional necrosis measurement and real-time PCR.

Main Results:

• Achieved 95% accuracy and a 0.99 Jaccard index for disease detection and quantification.
• The combined Cascade R-CNN and U-Net model effectively detected small, irregular lesions.
• Automated measurements showed high correlation with classical and molecular methods.

Conclusions:

• The developed deep learning model provides accurate and automated quantification of net blotch disease.
• This tool can be integrated into automated systems for disease monitoring and biocontrol agent screening.