Wheat Leaf Disease Classification using EfficientNet B3 Pre-Trained Architecture

Abstract

One of the most significant and widely grown cereal crops worldwide, wheat provides a substantial section of the world’s population with their primary source of sustenance. Due to its numerous benefits, wheat production and health significantly impact global population nutrition, economic stability, and food security. Maintaining global food security and promoting agricultural resistance depend on wheat agriculture’s long-term viability and productivity. The Indian Wheat Board assists wheat farmers in increasing wheat output by preventing several diseases in the wheat plant. A variety of climatic factors and other factors causes these illnesses. Early detection of wheat disease may reduce the damage to wheat leaves. This work aimed to use deep learning algorithms to tackle the problem of efficiently managing several illnesses that harm wheat leaves. This proposed work demonstrated the algorithms' potential for automating disease diagnosis and establishing the way for more precise and accurate detection techniques. The EfficientNetB3 network model architecture is used in this paper to classify photos of wheat leaves into three categories: healthy, septoria, and stripe-rust. The 407 images of wheat leaves from the collection were used. With a batch size of 30, the default learning rate, and 40 training rounds, the EfficientNetB3 model was used. The experimentation phase’s findings demonstrate the effectiveness of the suggested strategy. The EfficientNetB3 model performed well, obtaining a remarkable accuracy rate of 95% in categorizing the various kinds of wheat leaf conditions.