A Deep Learning Approach of Vehicle Multitarget Detection from Traffic Video

Table 3

Statistics of different model results for blocking flow sample training.

frame sample	YOLO 9000			YOLO-voc			YOLO v3			YOLO-vocRV
	free flow	225	214	2.5%	225	222	0.7%	225	174	12.8%	225	219	1.4%
synchronous flow	547	518	2.7%	547	479	6.6%	547	431	11.9%	547	523	2.2%
blocking flow	1050	955	4.7%	1050	970	4.0%	1050	799	13.6%	1050	992	3.7%