Table 3.

Robustness validation of the model

Model and Methods of Image corruption	Internal Test Set				External Test Set
	P@.5iou	R@.5iou	F1@.5iou	mAP@.5	P@.5iou	R@.5iou	F1@.5iou	mAP@.5
YOLOv8l	0.977	0.943	0.960	0.977	0.825	0.743	0.780	0.814
Gaussian noise	0.956	0.911	0.933	0.959	0.805	0.711	0.755	0.795
Shot noise	0.948	0.911	0.929	0.960	0.792	0.710	0.749	0.797
Impulse noise	0.954	0.889	0.920	0.949	0.805	0.684	0.739	0.787
Defocus blur	0.965	0.932	0.948	0.971	0.812	0.735	0.771	0.807
Zoom blur	0.810	0.758	0.783	0.837	0.680	0.587	0.630	0.705
Motion blur	0.966	0.966	0.966	0.969	0.818	0.732	0.772	0.803
Fog	0.945	0.661	0.778	0.830	0.796	0.493	0.610	0.689
Brightness +	0.977	0.943	0.960	0.977	0.824	0.743	0.780	0.815
Brightness -	0.956	0.932	0.944	0.971	0.804	0.731	0.766	0.803

P@.5iou = Precision with an Intersection over Union threshold of 0.5; R@.5iou = Recall with an Intersection over Union threshold of 0.5; F1@.5iou = F1 Score with an Intersection over Union threshold of 0.5; mAP@.5 = Mean Average Precision with an Intersection over Union threshold of 0.5.