Integrated YOLOv4 Deep Learning Pretrained Model for Accurate Estimation of Wheat Rust Disease Severity

Wheat is one of the most important crops globally, providing a significant portion of the world's food supply. However, wheat diseases, such as stripe rust, can significantly impact crop yield and quality. This study presents a novel approach for the multi-classification of Wheat Stripe Rust (WSR) disease into 5 different severity levels using the You Only Look Once version 4 (YOLOv4) model. The model was trained on a dataset of 5000 images and achieved a mean average precision (mAP) score of 0.85 and overall accuracy of 85.98%, demonstrating high accuracy and performance across all severity levels. This study also visualized the predicted bounding boxes to provide insights into how the model detected and classified different severity levels of WSR disease. Proposed findings have important implications for crop management and disease control strategies, as early detection and accurate classification of plant diseases can help to minimize crop losses and increase yields. The YOLOv4 model could be used as a powerful tool for agricultural decision-making, enabling farmers and agronomists to quickly and accurately identify plant diseases and take appropriate action. Overall, the study demonstrates the potential of deep learning (DL) models for accurate and efficient multi-classification of plant diseases and highlights the importance of continued research in this area to develop more sophisticated and effective models.