HYBRID DEEP LEARNING APPROACH FOR DETECTION OF BANANA PLANT DISEASES

Abstract

This study proposes a novel framework for banana plant disease detection by integrating machine learning and Deep learning techniques. The image dataset of banana plant collected from the agriculture fields in Jalgaon region of Maharashtra. The features from the images are extracted using transfer learning models such as EfficientNetB0, ResNet50, MobileNet and VGG16. These features are used for the classification of the banana plant whether it is healthy or diseased. The classification using extracted features performed using machine learning model such as logistic regression (LR), Support vector machine (SVM), Random Forest (RF), K-Nearest Neighbor (KNN) and XGBoost. Experimental results indicate that for MobileNet features LR achieved peak classification accuracy of 92.51%, while KNN demonstrated superior performance with both EfficientNetB0 and VGG16 feature extractors. ResNet50 based features also yielded optimal results with KNN of 91.68% accuracy. Comparative analysis revealed consistent performance by LR and SVM, with RF and XGBoost showing marginally lower efficacy. The findings validate that hybrid deep learning and machine learning architectures can significantly enhance plant disease prediction accuracy in the field of agriculture.