Leveraging the performance of Deep Learning Models for Corn Leaf Disease Diagnosis using DenseNet201 and Xception

Abstract

Plant diseases cause large output decreases and financial losses, making them a major barrier to global food security, especially in developing nations like Nepal, where corn is a staple crop. Early and accurate detection is critical to mitigating these impacts and improving crop management. Traditional diagnostic methods, reliant on manual inspection, are often time-consuming, subjective, and impractical for large-scale agricultural applications. This paper explores the automatic categorization of corn leaf diseases using deep learning-driven Convolutional Neural Networks (CNNs), specifically DenseNet 201 and Xception architectures. Convolutional layers in these models learn and extract distinctive features automatically from images, enabling accurate and efficient classification of corn disease types. A freely accessible dataset comprising images of both healthy and diseased corn leaves was utilized, with data augmentation strategies used to enhance model generalization and robustness. Experimental results demonstrate that DenseNet201 achieved a test accuracy of 98.69%, outperforming Xception, which attained 96.61%. These results demonstrate the highlights of CNN-based approaches for scalable, non-invasive, and accurate disease detection in corn crops. The proposed method offers a viable tool to support precision agriculture and contribute to enhancing global food security.