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

Table 1

CNN architecture.

Layers		Input size	Output size
Convolution	conv, str.2
Pooling	max pool, str.2
Dense block 1	conv
	conv
Transition 1	conv
	avr. pool., str.2
Dense block 2	conv
	conv
Transition 2	conv
	avr. pool., str.2
Dense block 3	conv
	conv
Transition 3	conv
	avr. pool., str.2
Dense block 4	conv
	conv
Custom embedder	global avr. pool
	Batch normalisation
	Dropout (50%)
	Fully connected layer
	Batch normalisation
Classifier	Output layer