Safety Helmet-Wearing Detection System for Manufacturing Workshop Based on Improved YOLOv7

Table 3

Performance of different algorithms on three data sets.

Data set	Methods	(%)	(%)	mAP (%)	F1 (%)
Helmet-head	Faster RCNN	92.91	90.81	94.04	91.85
	YOLOv5l	94.83	93.75	96.51	94.29
	YOLOv5m	94.64	93.07	96.24	93.85
	YOLOv5s	94.41	92.03	95.79	93.20
	YOLOv5x	94.74	94.15	96.62	94.44
	YOLOv7-tiny	92.94	90.86	95.09	91.89
	YOLOv7-d6	92.40	88.03	93.72	90.16
	YOLOv7-e6	94.11	92.22	96.08	93.16
	YOLOv7-e6e	93.22	90.73	95.08	91.96
	YOLOv7-w6	92.48	89.34	94.31	90.88
	YOLOv7	94.71	94.02	97.15	94.36
	YOLOv7-x	94.38	94.52	97.26	94.45
	YOLOv7_ours	94.61	94.76	97.54	94.68
Helmet-data	Faster RCNN	88.65	0.8562	90.19	87.11
	YOLOv5l	91.64	88.14	94.31	89.86
	YOLOv5m	93.05	86.42	91.89	89.61
	YOLOv5s	92.11	86.51	91.76	89.22
	YOLOv5x	92.02	87.66	92.29	89.79
	YOLOv7-tiny	90.83	88.53	92.94	89.67
	YOLOv7-d6	91.55	86.95	92.21	89.19
	YOLOv7-e6	92.38	85.36	91.73	88.73
	YOLOv7-e6e	91.66	87.73	92.77	89.65
	YOLOv7-w6	90.22	87.21	91.97	88.69
	YOLOv7	92.73	89.59	94.06	91.13
	YOLOv7-x	93.19	89.82	94.36	91.47
	YOLOv7_ours	92.73	90.47	94.76	91.59
Helmet	Faster RCNN	81.35	0.7373	80.59	77.35
	YOLOv5l	85.24	72.75	79.01	78.50
	YOLOv5m	89.72	69.14	77.69	78.10
	YOLOv5s	86.32	69.67	77.51	77.11
	YOLOv5x	85.36	74.01	79.41	79.28
	YOLOv7-tiny	85.66	71.71	79.43	78.07
	YOLOv7-d6	83.76	81.22	85.51	82.47
	YOLOv7-e6	87.71	77.85	85.00	82.49
	YOLOv7-e6e	86.62	78.13	84.14	82.16
	YOLOv7-w6	85.33	79.27	83.17	82.19
	YOLOv7	87.72	77.88	85.19	82.51
	YOLOv7-x	86.11	76.45	83.97	80.99
	YOLOv7_ours	87.54	80.40	85.98	83.82