Pavement Disease Detection through Improved YOLOv5s Neural Network

Table 2

Comparison of different model performance metrics.

Model	Pre-T (ms)	Inference (ms)	NMS (ms)	FPS	Parameter (M)	FLOPs (G)
YOLOv3	4.0	311	388	3.2	40.585	24.5
YOLOv3 + Four scale	4.0	331	388	3.0	55.792	33.7
YOLOv5s	2.7	223	301	4.4	6.900	16.5
YOLOv5s + Ghostbackbone	2.7	128	301	7.7	4.866	9.5
YOLOv5s + ShuffleNetv2	2.7	109	248	9.0	0.811	1.6
YOLOv5s + MobileNetv3	2.7	92	231	10.6	1.417	0.8
Our Ghost-YOLOv5s	2.6	72	301	12.5	2.104	2.2