Hierarchy in Disease Severity Classification in Banana Leaves through the Integration of Convolutional Neural Network and Decision Tree Models
Raghav Jain, Vinay Kukreja, Saumitra Chattopadhyay, Aditya Verma, Rishabh Sharma
Abstract
Accurate and timely identification of banana leaf diseases is very important for effective disease control, because such diseases are some of the most serious threats to agricultural productivity. Banana disease severity classification across six levels (Level 0 to Level 5) using machine learning models is the novel feature of this study. With about 900 collected samples (including images captured for diseases at various stages), and manual annotation done as a rigorous process of preprocessing, customized versions Convolutional Neural Networks (CNNs) were constructed with Decision Tree. Their ability to differentiate nuanced variations in severity of the disease was demonstrated by models' excellent overall accuracy 96.04%. Balanced classification capabilities of the models were reflected in performance evaluation metrics such as precision, recall and F1-score. Comparative studies showed the developed models to be superior, so that they may become integral parts of early disease detection and intervention in banana cultivation. Though limited in scope, this paper has value for the agricultural sciences. It presents a framework that can accurately classify disease severity and provides basic material upon which more effective crop management strategies may be built or higher yielding standards established.