Fusing Two Kinds of Virtual Samples for Small Sample Face Recognition

Table 1

Classification error rates of the COVNFR method.

Database	The number of training samples	1	2	3
	CRC-original	0.3194	0.1656	0.1393
ORL	CRC-virtual training samples only	0.9306	0.9313	0.9643
	CRC-COVNFR	0.3778	0.2719	0.2750
	CRC-original	0.3733	0.1556	0.1083
Yale	CRC-virtual training samples only	0.9000	0.9333	0.9667
	CRC-COVNFR	0.3533	0.1259	0.1000
	CRC-original	0.5567	0.4160	0.5563
FERET	CRC-virtual training samples only	0.9592	0.8810	0.7937
	CRC-COVNFR	0.6200	0.5290	0.6525
	CRC-original	0.8437	0.7660	0.7641
FLW	CRC-virtual training samples only	0.9470	0.9608	0.9269
	CRC-COVNFR	0.9031	0.8314	0.8306