Table 4.
Descriptions of initial parameters for training of various CNN models (“# of” represents “the number of”).
| Dataset | # of Output Class | CNN Model | Max. # of Iteration (Epoch) | Mini-Batch Size | Learning Rate | Momentum/Gamma | |
|---|---|---|---|---|---|---|---|
| SDU-DB | Finger-vein | 2 | VGG Net-16 | 7651 (10) |
64 | 0.001 | 0.9/0.1 |
| ResNet-50 | 24,486 (10) |
20 | |||||
| ResNet-101 | 32,648 (10) |
15 | |||||
| Finger shape | 318 | VGG Net-16 | 36,073 (10) |
64 | |||
| ResNet-50 | 115,434 (10) |
20 | |||||
| ResNet-101 | 153,912 (10) |
15 | |||||
| PolyU-DB | Finger-vein | 2 | VGG Net-16 | 3753 (10) |
64 | ||
| ResNet-50 | 12,012 (10) |
20 | |||||
| ResNet-101 | 16,016 (10) |
15 | |||||
| Finger shape | 156 | VGG Net-16 | 17696 (10) |
64 | |||
| ResNet-50 | 56,628 (10) |
20 | |||||
| ResNet-101 | 75,504 (10) |
15 | |||||