Hybrid ensemble - deep transfer model for early cassava leaf disease classification.

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Cassava is a vital carbohydrate food source in Africa and Asia.
• Cassava leaf diseases significantly impact crop production.
• Accurate and early disease detection is crucial for crop yield and food security.

Purpose of the Study:

• To develop an effective approach for early cassava leaf disease detection.
• To address the challenge of imbalanced datasets in disease classification.
• To improve the accuracy of automated cassava disease identification using deep learning.

Main Methods:

• Developed a hybrid Ensemble-deep transfer model approach.
• Utilized data augmentation techniques to balance the dataset.
• Created three novel hybrid models: Ensemble(InceptionV3+DenseNet-BC-121-32 + Xception), Ensemble(ResNet50V2+DenseNet-BC-121-32), and Ensemble(ResNet50V2+ResNet50).

Main Results:

• The proposed hybrid ensemble models demonstrated high performance.
• Achieved the highest accuracy of 88.83% for five-class classification.
• Obtained the highest accuracy of 97.89% for two-class classification using the Ensemble(InceptionV3+DenseNet-BC-121-32 + Xception) model.

Conclusions:

• The hybrid Ensemble-deep transfer model approach is effective for early cassava leaf disease detection.
• Data augmentation and ensemble methods significantly enhance classification accuracy.
• The developed models offer a promising solution for automated plant disease diagnosis.