A Multibranch Object Detection Method for Traffic Scenes

Table 8

The comparison of average intersection over union (IOU) for various methods (%).

Model	Threshold
	0.1	0.15	0.2	0.25	0.3	0.35	0.4	0.45	0.5	0.55	0.6	0.65
RCNN [4]	72.45	71.12	68.67	66.89	63.21	61.72	59.56	56.01	53.29	50.67	46.59	41.21
Faster RCNN [9]	78.23	77.67	76.55	74.31	71.29	69.11	67.47	65.12	62.89	59.61	57.23	54.45
SSD [15]	84.67	82.58	80.69	78.97	76.29	75.17	73.90	70.55	68.22	65.14	62.25	58.30
Mask RCNN [13]	80.34	79.25	78.58	78.10	77.31	75.69	74.26	72.79	70.99	68.68	64.36	60.01
SINet [2]	88.90	88.08	82.29	81.81	80.22	76.21	73.33	68.68	66.43	64.42	63.99	60.68
YOLOv3 [17]	85.87	84.01	81.44	79.45	77.39	75.88	72.32	69.73	68.97	65.82	64.29	61.33
MBNet	89.25	87.27	84.45	82.66	79.99	77.11	74.77	71.38	68.20	66.23	65.01	61.58