TCM: Temporal Consistency Model for Head Detection in Complex Videos

Table 5

Speed versus accuracy of different detectors before and after employing the proposed TCM model.

	Detection models	Before TCM		After TCM
		mAP	Speed (FPS)	mAP	Speed (FPS)
Generic detectors	Faster R-CNN [15]	0.80	17	0.87	10
	YOLO [14]	0.72	50	0.82	43
	SSD [13]	0.75	52	0.81	47
Specific detectors	Context-CNN [16]	0.81	10	0.88	6
	FCHD [48]	0.76	12	0.83	8
	Faceness [29]	0.79	19	0.82	13