Finger-Vein Recognition Using Bidirectional Feature Extraction and Transfer Learning

Table 5

The recognition accuracy improvement of concatenated feature recognition models, where the last column is compared with the unidirectional models testing on A FV-USM and A FV-SIPL in Table 4.

Network	Database	Accuracy (%)	Improvement (%)
CNN (VGG19)	A and B FV-USM	98.00	1.73
CNN (ResNet50)		99.67	1.36
CNN (VGG19)	A and B FV-SIPL	99.07	0.07
CNN (ResNet50)		99.31	0.24