Deep learning-based phenotyping of lettuce diseases using Efficient-FBM-FRMNet for precision agriculture

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Lettuce (Lactuca sativa) is vulnerable to bacterial and fungal diseases, impacting crop yield and quality.
• Accurate and rapid disease identification is crucial for effective precision agriculture and sustainable crop management.

Purpose of the Study:

• To develop and evaluate Efficient-FBM-FRMNet, a novel deep learning framework for automated lettuce disease detection.
• To assess the model's performance against established Convolutional Neural Networks (CNNs) and its potential for real-world applications.

Main Methods:

• A modular deep learning framework, Efficient-FBM-FRMNet, was designed, integrating EfficientNetB4 with dilated convolutions, a Feature Bottleneck Module (FBM), a Reasoning Engine, and a Feature Refinement Module (FRM).
• The model was trained and validated on 2,813 lettuce leaf images across bacterial, fungal, and healthy classes using stratified 5-fold cross-validation.

Main Results:

• Efficient-FBM-FRMNet achieved a high overall accuracy of 97.5%, surpassing baseline CNNs like EfficientNetB4, ResNet50, and DenseNet121.
• The model demonstrated excellent precision (96.0%), recall (96.6%), and F1-score (97.0%), with statistically significant performance gains (p < 0.05).
• The framework is computationally efficient, featuring a small model size (8.2 MB) and fast inference time (23 ms).

Conclusions:

• Efficient-FBM-FRMNet offers a robust, accurate, and efficient solution for automated lettuce disease detection.
• The model's enhanced feature learning, interpretability, and stability make it suitable for deployment in precision agriculture systems, including greenhouse monitoring and UAV surveillance.
• This AI-driven approach contributes to sustainable crop management by enabling early and precise disease identification.