AUTOMATED DETECTION OF BANANA LEAF DISEASES USING DEEP LEARNING TECHNIQUES: A COMPARATIVE STUDY OF CNN ARCHITECTURES

Abstract

Abstract-Banana leaf diseases pose a significant threat to global banana production, impacting both crop yield and quality. Traditional methods of disease detection often rely on manual inspection, which can be time-consuming and subjective. This paper explores the application of deep learning techniques for the automatic detection and classification of banana leaf diseases, specifically focusing on diseases such as Black Sigatoka, Mosaic disease, Anthracnose, Xanthomonas wilt, and Fusarium wilt. Utilizing various convolutional neural network architectures—including VGG19, XceptionNet, ResNet50, DenseNet, SqueezeNet, and a generic CNN—we developed a robust framework for image classification based on leaf images captured in diverse field conditions. The dataset comprises a variety of images representing healthy and diseased leaves, annotated for training and validation purposes. We employed data augmentation techniques to enhance model performance and generalization. These experimental results demonstrate high accuracy and precision in detecting specific diseases, with certain models outperforming conventional methods. This study highlights the potential of deep learning as a powerful tool for disease monitoring and management in banana cultivation, contributing to more sustainable agricultural practices and improved food security.