Corn Disease Severity Estimation using U²Net Based Convolution Neural Network

This paper introduces a deep learning-based approach for estimating the severity of maize plant diseases through Convolutional Neural Networks (CNNs). Diseases like Northern Leaf Blight (NLB), Grey Leaf Spot (GLS), and Northern Leaf Spot (NLS) significantly hinder maize production, posing a risk to global food security. Traditional visual inspection methods, particularly in large-scale agriculture, often lack precision, leading to potential mismanagement and overuse of pesticides. This research explores the effectiveness of U<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>Net based image enhancement as an image preprocessing tool and using a lightweight MobileNet CNN models in accurately detecting and categorizing the severity of maize diseases from images, offering a scalable solution for disease management. The model achieves high accuracy of 94.15 % in classification as well as 94% accuracy in severity estimation in 5 levels of disease severity on the field images of Northern Leaf Spot (NLS) in CD&S dataset, highlighting its potential for integration into automated agricultural monitoring systems.