Automated Detection of Banana Fruit Diseases Using Deep Learning

Abstract

The use of deep learning in the automation of banana fruit disease diagnosis is a new technique that has been used to deal with a major problem in agriculture and food safety. Little or poorly identified diseases can cause significant losses after harvest, low prices at the market, and dissatisfaction of customers. In this study, it is proposed to apply the most effective deep learning methods to banana fruit images to detect the various types of diseases and distinguish between diseased and healthy fruits. Various convolutional neural network (CNN) models were investigated, but the DenseNet121 model presented the most beneficial results. In this model the accuracy was impressive (99.12) on the first dataset and 98.42 on the second dataset in determining the disease types of banana fruits. The suggested method deals with the requirement of scalable, fast, and accurate solutions in agricultural supply chains, in particular, in the regions where human inspection is either unreliable or absent. Despite the existing shortcomings, including the fluctuating climate, disease presentation variations, and absence of data with annotations, the study demonstrates the prospective of AI-based systems to transform the assessment of agricultural illnesses. Our solution provides a cost effective and user friendly application with which farmers, retailers and consumers can make an informed decision. This study improves timely detection and management of disease as well as timely response thus reducing agricultural losses and increasing food distribution channels. The combination of AI-based technologies and mobile technology would enhance the diagnosis of diseases in isolated farmlands. This technology can be used to enhance food security throughout the world and in the long-term sustainable agricultural processes.