A Comparative Study of Cross-Device Finger Vein Recognition Using Classical and Deep Learning Approaches

Table 6

EER and FNMR1000 performances in the percentage of Miura method, CNN, and CAE with different device pairs.

Miura
	UTFV		ZkTeco		IDIAP		PFV_L		PFV_C		NTNU
	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000
UTFV	0.57	1.52	10.4	50.19	25.0	82.67	47.1	97.54	49.8	99.83	47.7	99.13
ZkTeco	15.1	40.91	5.13	24.63	35.0	88.16	45.4	97.78	34.4	98.04	49.4	99.83
IDIAP	36.4	88.44	29.1	66.56	16.2	44.74	37.6	89.12	40.3	94.58	39.9	99.12
PFV_L	48.7	97.54	44.4	99.81	44.7	99.92	9.52	34.52	42.3	87.98	49.8	100.0
PFV_C	49.8	99.83	48.5	98.37	37.4	97.92	43.1	90.16	22.6	58.33	46.8	99.67
NTNU	49.3	99.82	49.9	99.91	48.5	100.0	46.8	100.0	49.5	100.0	32.6	76.30
CNN
	UTFV		ZkTeco		IDIAP		PFV_L		PFV_C		NTNU
	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000
UTFV	12.7	95.19	24.4	93.74	33.4	97.58	42.4	99.64	42.2	99.42	47.8	99.88
ZkTeco	24.4	93.74	10.6	85.02	46.4	99.43	41.9	97.71	34.2	99.54	49.9	99.97
IDIAP	33.4	97.58	46.4	99.43	23.2	99.57	43.3	99.41	48.5	100.0	49.5	99.64
PFV_L	42.4	99.94	41.9	97.71	43.3	99.41	25.0	97.13	39.6	99.96	48.1	99.50
PFV_C	42.2	94.42	34.2	99.54	48.5	100.0	39.6	99.96	27.4	96.62	49.5	99.49
NTNU	47.8	99.88	49.9	99.97	49.5	99.64	48.1	99.50	49.5	99.49	34.3	98.93
CAE
	UTFV		ZkTeco		IDIAP		PFV_L		PFV_C		NTNU
	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000	EER	FMR1000
UTFV	1.25	3.03	7.95	18.47	12.3	55.18	34.8	97.54	40.1	98.83	48.2	99.88
ZkTeco	9.85	17.42	2.41	12.96	21.3	75.33	28.9	96.67	25.8	88.24	49.4	100.0
IDIAP	10.3	49.66	22.3	75.00	14.5	49.78	37.0	97.45	47.6	94.72	49.1	99.56
PFV_L	26.8	95.45	39.4	99.07	48.3	99.19	8.62	34.52	42.1	94.54	48.5	100.0
PFV_C	42.8	98.67	48.5	98.53	42.9	99.97	29.8	73.77	14.3	36.90	49.3	99.83
NTNU	46.9	99.91	49.9	100.0	48.2	99.89	49.4	100.0	49.1	99.50	40.7	95.64