Enhancing Banana Leaf Spot Disease Classification using Dense Mobilenet V2

Bananas are an essential economic crop, generating substantial revenue and employment in tropical and subtropical regions worldwide. However, banana production faces serious challenges due to leaf spot diseases like Sigatoka, Cordana and Pestalotiopsis which reduce yield and increase production costs. To combat these diseases, deep learning methods are being used to enhance disease detection and management, thereby improving the overall profitability and sustainability of banana farming. This study proposes a hybrid deep learning model called DenseMobilenetV2 which is the combination of the very reliable DenseNetl21 and the swift MobilenetV2 and trained and verified on a dataset comprising 937 images. Preprocessing methods like image invention, image augmentation, and outlier analysis were deployed to reduce outliers and increase the diversity of the dataset. After training it for 25 epochs, 98.42% accuracy was achieved alongside the 1.00 precision score was attained. Fl-score and recall accuracy are 0.9987 and 0.9974 respectively which indicates the robustness of the proposed model. The lightweight nature of the model due to the presence of the MobilenetV2 will help the model to be integrated into real-life applications helping novice and veteran farmers.