A Comparative Study on Enhancing Wheat Leaf Disease Detection and Classification Through Deep Learning Approaches

Abstract

Many leaf diseases have significant effects on yield and quality and wheat is an important crop contributing to food security globally. Accurate and timely diagnosis of these diseases is important for the sustainable use of agriculture. This study assesses the effectiveness of deep learning (DL) technique using ResNet for wheat leaves identification and classification. The proposed architecture aims to address the drawbacks of the conventional Convolutional Neural Networks (CNNs) by using ResNet50, ResNet101 and ResNet152. A variety of parameters are used to assess prediction models on benchmark datasets. The research suggests that ResNet50 beats other classes of ResNets with a classification accuracy of 90%. Using improved learning rates and optimized feature extraction layers, the network is able to successfully differentiate between healthy and unhealthy wheat leaves. For agricultural monitoring systems, this research reveals that ResNet models have great potential to accurately identify diseases in real-time.