Table 5.

Quantitative comparison of fracture detection when the input image size is 640.

Model	mAP${}^{\mathit{val}}$ 50	mAP${}^{\mathit{val}}$ 50–95	Speed CPU Intel Core i5 (ms)	Speed GPU RTX 3080Ti (ms)	PARAMS (M)	FLOPs (B)
YOLOv5n	0.589	0.339	\	2.8	1.77	4.2
YOLOv8n	0.601	0.374	67.4	2.9	3.01	8.1
Ours	0.605	0.379	111.3	3.4	3.01	8.2
YOLOv5s	0.601	0.357	\	3.3	7.03	15.8
YOLOv8s	0.604	0.383	191.5	4.3	11.13	28.5
Ours	0.612	0.392	285.1	4.9	11.13	28.7
YOLOv5m	0.613	0.371	\	4.0	20.89	48.0
YOLOv8m	0.621	0.403	536.4	5.5	25.84	78.7
Ours	0.629	0.404	685.9	5.1	25.84	78.7
YOLOv5l	0.620	0.379	\	5.6	46.15	107.8
YOLOv8l	0.624	0.403	1006.3	7.4	43.61	164.9
Ours	0.637	0.406	1370.8	7.2	43.61	164.9

Speed means the total time of validate per image, and the total time includes the preprocessing, inference, and post-processing time.