Table 1.

Characteristics of the architectures of the studied models.

Model name	Depth	Type of pooling	Mask size	# channels of the conv. layers	# Layers	# Parameters
CUnet1	4	maximum $\left(2\times 2\right)$	3 × 3	[16, 32, 64, 128, 256]	82	2,167,636
CorneaNet	4	average $\left(2\times 2\right)$	3 × 3	[16, 32, 64, 128, 256]	82	2,167,636
CUnet3	3	maximum $\left(2\times 2\right)$	3 × 3	[8, 16, 32, 64]	64	136,980
CUnet4	3	average $\left(2\times 2\right)$	3 × 3	[8, 16, 32, 64]	64	136,980
CUnet5	3	average $\left(2\times 2\right)$	5 × 5	[32, 64, 128, 256]	247	23,871,732