Detection of the Metcalfa pruinosa (Hemiptera: Flatidae) pest on the Jujube plant (Ziziphus jujuba) using a sequence of YOLOv5 models

Abstract

This study aimed to detect the adult of the pest Metcalfa pruinosa observed on jujube plants using the YOLOv5 algorithm's v5s, v5m, and v5l models. Performance metrics, including box_loss, obj_loss, precision, recall, mAP_0.5, and mAP_0.5:0.95, were observed in the research. In the YOLOv5s model, the box_loss and obj_loss performance metrics were found to be the highest, with values of 0.02858 and 0.0055256, respectively. In the YOLOv5m model, the recall performance metric was identified as the highest, with a value of 0.98127. In the YOLOv5l model, precision, mAP_0.5, and mAP_0.5:0.95 performance metrics were identified as the highest, with values of 0.98122, 0.99500, and 0.67864, respectively. Consequently, the YOLOv5l model exhibits higher precision compared to others. It is believed that the YOLOv5l model is sufficient for the detection of the Metcalfa pruinosa pest.