Classification of Banana Leaf Diseases Using Image-Based Convolutional Neural Network Method

Abstract

Banana leaf diseases have a significant impact on the productivity and quality of banana plants, especially in tropical areas. Early and accurate detection of the type of disease is very important to prevent wider spread. This research aims to develop an image-based banana leaf disease classification model using the Convolutional Neural Network (CNN) method. The banana leaf image dataset containing various types of diseases is processed through data augmentation techniques to increase the accuracy and robustness of the model. This research compares three CNN models GlobalAveragePooling, Averagepooling2D, and Flatten. The research results show that the GlobalAveragePooling model achieved the highest accuracy of 93.62%, followed by AveragePooling2D with an accuracy of 90.96%, and Flatten with an accuracy of 88.83%. It is hoped that the implementation of this model can help farmers detect banana leaf diseases quickly and precisely, support more effective preventive measures, and increase the productivity of banana plants.

Keywords: banana leaf disease, Convolutional Neural Network, disease detection, image classification, data augmentation