Deep learning based ensemble model for accurate tomato leaf disease classification by leveraging ResNet50 and...

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Tomato cultivation is crucial for global food security, but diseases significantly reduce yield and quality.
• Accurate and early disease diagnosis is essential for mitigating crop losses and financial impacts.

Purpose of the Study:

• To develop and evaluate a deep learning-based ensemble model for automated tomato leaf disease classification.
• To combine the strengths of MobileNetV2 and ResNet50 for enhanced feature extraction and classification accuracy.

Main Methods:

• An ensemble model was created by integrating MobileNetV2 and ResNet50, with modified output layers for improved feature extraction.
• A dataset of 11,000 annotated images covering 10 tomato disease categories and healthy leaves was utilized.
• Data preprocessing involved image resizing, splitting (80% train, 10% test, 10% validation), and model training.

Main Results:

• The proposed ensemble model achieved a remarkable test accuracy of 99.91%.
• The model demonstrated high performance metrics: 99.92% precision, 99.90% recall, and a 99.91% F1-score.
• A confusion matrix confirmed near-flawless classification with minimal misclassifications across all disease categories.

Conclusions:

• Deep learning offers a scalable and accurate solution for automating tomato disease diagnosis.
• The developed model can significantly improve crop health monitoring and reduce economic losses in agriculture.
• This approach supports precision agriculture and sustainable farming practices through early disease intervention.