Automated severity level estimation of wheat rust using an EfficientNet-CBAM hybrid model

Area of Science:

• Agricultural Science
• Plant Pathology
• Computer Vision

Background:

• Wheat rust diseases cause significant global crop losses, impacting yield and quality.
• Accurate and timely disease severity estimation is vital for effective agricultural management.
• Early detection of wheat rust enables prompt intervention to mitigate crop damage.

Purpose of the Study:

• To develop an automated model for accurate wheat rust severity stage estimation.
• To improve feature extraction in deep learning models for plant disease identification.
• To create a practical tool for real-time disease assessment in the field.

Main Methods:

• Utilized the EfficientNet-B0 architecture integrated with a convolutional Block Attention Module (CBAM).
• Trained the model on a diverse image dataset of wheat rust (stripe, stem, leaf) and healthy plants.
• Categorized disease severity into four stages: healthy, low, medium, and high.

Main Results:

• The proposed model achieved a high training accuracy of 99.51% and testing accuracy of 96.68%.
• Demonstrated superior performance compared to other state-of-the-art Convolutional Neural Network (CNN) models.
• Developed an Android application for real-time wheat rust severity classification.

Conclusions:

• The developed EfficientNet-B0 model with CBAM effectively automates wheat rust severity estimation.
• The system offers a promising solution for early disease detection and management in wheat cultivation.
• The mobile application provides a user-friendly tool for farmers to monitor crop health.