Table 9.

Evaluation metrics for KITTI point cloud data compressed at different degrees, converted with the Delaunay Triangulation and evaluated using the best YOLOv5m model, trained over the course of 200 epochs with a batch size of 20.

Class	Precision				Recall				mAP
	Raw	G1	G2	G3	Raw	G1	G2	G3	Raw	G1	G2	G3
Car	0.855	0.879	0.882	0.842	0.859	0.848	0.796	0.759	0.917	0.916	0.888	0.849
Pedestrian	0.893	0.922	0.894	0.594	0.718	0.688	0.658	0.617	0.791	0.789	0.747	0.605
Cyclist	0.842	0.847	0.857	0.805	0.713	0.693	0.634	0.468	0.779	0.774	0.748	0.604
Average	0.863	0.883	0.878	0.747	0.763	0.743	0.696	0.614	0.829	0.827	0.794	0.686