Table 1 Detection accuracy for the three... | Emerald Publishing

Table 1

Detection accuracy for the three different subsets in the DeepStreets for both our method and the state-of-the-art XceptionNet method in [1]. Results are reported for different video qualities (Raw, HQ and LQ).

No. train c	City vid						Citywcvid						Kittyvid
	Raw		HQ		LQ		Raw		HQ		LQ		Raw		HQ		LQ
	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept
2	96.9	50.0	97.5	50.0	95.8	50.0	96.7	50.0	98.6	50.0	99.0	50.0	80.0	50.0	75.0	50.0	67.8	50.0
4	97.2	50.0	98.5	50.0	98.3	50.0	98.5	55.0	98.9	67.0	99.6	52.0	81.1	50.0	76.2	50.0	71.0	50.0
8	98.5	88.0	98.4	78.0	99.5	84.0	98.5	70.0	99.8	71.0	99.7	84.0	82.1	99.0	81.2	100	72.1	100
16	98.6	90.0	99.0	94.0	99.5	81.0	99.7	90.0	100	87.0	99.8	92.0	85.0	100	82.8	100	74.2	100
32	98.8	97.0	99.1	66.0	99.7	89.0	99.9	96.0	100	97.0	99.9	97.0	86.8	100	86.8	100	76.6	85.0
64	99.4	87.0	99.4	100	99.7	99.0	100	100	100	100	100	86.0	87.9	100	85.3	100	79.5	99.0
128	99.0	100	100	100	100	96.0	100	100	100	100	100	100	91.6	100	91.3	100	81.5	100

No. train c	City vid						Citywcvid						Kittyvid
	Raw		HQ		LQ		Raw		HQ		LQ		Raw		HQ		LQ
	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept	F-SVM	Xcept
2	96.9	50.0	97.5	50.0	95.8	50.0	96.7	50.0	98.6	50.0	99.0	50.0	80.0	50.0	75.0	50.0	67.8	50.0
4	97.2	50.0	98.5	50.0	98.3	50.0	98.5	55.0	98.9	67.0	99.6	52.0	81.1	50.0	76.2	50.0	71.0	50.0
8	98.5	88.0	98.4	78.0	99.5	84.0	98.5	70.0	99.8	71.0	99.7	84.0	82.1	99.0	81.2	100	72.1	100
16	98.6	90.0	99.0	94.0	99.5	81.0	99.7	90.0	100	87.0	99.8	92.0	85.0	100	82.8	100	74.2	100
32	98.8	97.0	99.1	66.0	99.7	89.0	99.9	96.0	100	97.0	99.9	97.0	86.8	100	86.8	100	76.6	85.0
64	99.4	87.0	99.4	100	99.7	99.0	100	100	100	100	100	86.0	87.9	100	85.3	100	79.5	99.0
128	99.0	100	100	100	100	96.0	100	100	100	100	100	100	91.6	100	91.3	100	81.5	100