Enhancing wheat yellow rust detection through modified deep learning approach

The study reports the effectiveness of machine learning and deep learning (ML/DL) algorithms in detecting yellow rust disease in wheat, keeping in view that early stage detection is crucial for minimizing the crop yield loss. Various traditional ML algorithms including Support Vector Machine (SVM), Decision Tree (DT), k-Nearest Neighbor KNN, Naïve Bayes (NB), Random Forest (RF) and Sequential Convolutional Neural Network (CNN) are utilized to access the plant disease detection efficiency. In contrast a modified version of CNN (MCNN), integrated with SVM principles, is proposed to enhance the performance of standard CNN. A newly established dataset named Yellow Rust 2022–2023 (YR-22/23) is used to estimate and compare the performance of considered ML/DL algorithms. Additionally, a benchmarked dataset named YellowRust-19 is considered for cross validation of ML/DL algorithms. The result indicates that the MCNN outperforms the standard algorithms (CNN) in terms of accuracy (1.2 %) metrics and is highly time efficient. Our proposed modified CNN method attained prediction accuracy of about 98 % for detection of yellow rust of wheat. Result highlights that merger of ML/DL approaches holds great promises to improve the overall efficiency of plant disease detection. • Proposed a novel MCNN architecture by integrating SVM principles to enhance wheat yellow rust detection using CNN. • Explored ML/DL approaches for wheat yellow rust detection on two datasets (YR–22/23 & YellowRust–19). • Conducted a detailed comparative analysis demonstrating the superiority of the MCNN over state–of–the–art methods.