Integrating YOLOv5 and Pretrained Models to Enhance Wheat Leaf Rust Disease Recognition

To ensure consistent and dependable wheat production and food security, effective monitoring of leaf rust is essential which will increase the wheat grain yield quality. The wheat diseases are recognized by either experienced analyzers or farmers and this process is too laborious. To solve these issues, a deep learning-based YOLOV5 model along with pre-trained models has been employed for wheat leaf rust disease recognition in this paper. At first, 400 images of wheat in total have been gathered from secondary sources. The YOLOV5 model uses several hyperparameters to generate the wheat leaves masks. Once, the wheat leaves masks have been obtained, the binary classification of the wheat leaf is needed. The binary classification of wheat healthy leave and wheat leaf rust diseased has been performed by VGG16 and VGG19 pre-trained models. The YOLOV5 model has the highest precision, recall, and mAP of 97.8%, 97.3%, and 98.6%. In terms of performance metrics, the VGG16 model produces a higher F1 score (92.89%) than the VGG19 model (86.52%). This study demonstrates the potential of deep learning-based models in the identification and management of wheat leaf rust disease, which can lead to significant improvements in wheat production and food security.