Finger Vein Verification on Different Datasets Based on Deep Learning with Triplet Loss

Table 5

The accuracy, precision, recall, ROC AUC, and best distance of corresponding training-validation sets in ResNet18 model. The values at .

Training-validation	Accuracy	Precision	Recall	ROC AUC	Best distances
FV-USM-HKPU; (F-H)				0.9965		0.8904
FV-USM-SDUMLA-HMT; (F-S)				0.9888		0.7206
HKPU-FV-USM; (H-F)				0.9973		0.9177
HKPU-SDUMLA-HMT; (H-S)				0.9927		0.8027
SDUMLA-HMT-FV-USM; (S-F)				0.9975		0.8884
SDUMLA-HMT-HKPU; (S-H)				0.9958		0.8503