Wheat Leaf Disease Detection Using ResNet50

Abstract

One of the most significant cereal crops in the world is wheat and it plays a major part in food security in the world [1][2]. The issue is that the production of wheat is also significantly affected by several leaf diseases that can cause significant losses of yields, Leaf Rust, Fusarium Head Blight, and Tan Spot, which could not be identified at an initial level [1][5]. The traditional disease detection techniques involve manual discovery by the experts which is time consuming, subjective and non-scalable [6][9]. In the recent years, the artificial intelligence enabled by the development of deep learning and computer vision has enabled automated and trustworthy diagnosis of diseases using the examination of images [2][7]. In this research paper, an automated wheat leaf disease detection system named DL-WheatNet is proposed, which utilizes a ResNet50 deep convolutional neural network with transfer learning [1][7]. The method suggested is accurate in classifying the images of wheat leaves into several categories of diseases and the healthy classification. The findings of the experiment reveal that the given approach is more accurate and resilient than the existing machine learning algorithms and simple CNN models. The system can be implemented as a practical decision-making tool among the farmers and agricultural experts.