Implementation and Performance Evaluation of YOLOv8 for Wheat Fungal Disease Detection

Abstract

Among the many difficulties in producing wheat, a vital food source for the world, there are biotic variables like fungal diseases, which drastically lower crop output. Around the world, biotrophic fungi particularly, leaf rust, powdery mildew, and yellow rusts-have become a constant threat to wheat production. Traditional methods of identification and mitigation are labor-intensive, slow, and imprecise, limiting therefore early detection and intervention of wheat diseases is essential. Convolutional Neural Networks (CNNs), are the recent development in deep learning techniques, have shown promising disease detection solutions. Among these, the YOLO (You Only Look Once) models-particularly YOLOv8-have shown remarkable efficiency in identifying and categorizing wheat diseases like leaf rust, yellow rust, and powdery mildew. In this study, the YOLOv8 model was trained on a wheat image dataset, and achieved a mean average precision (mAP) of 0.99. The outcomes demonstrate the model’s reliability in identifying diseases under a range of circumstances, emphasizing the significance of maximizing training time to prevent overfitting. To further improve model performance, future studies should investigate data augmentation, hyperparameter adjustment, and real-time deployment in agricultural settings. The study demonstrates how deep learning models can enhance crop monitoring, disease control, and eventually agricultural output.