Table 1.

Basic architecture and configuration parameters of both AlexNet and VGGNet, where Conv. means convolutional, FC means fully-connected, SGDM means stochastic gradient descent with momentum, and cross entr. means cross entropy.

	AlexNet		VGGNet
	Original	Fine-Tuned	Original	Fine-Tuned
No. of training images	>1 million	329–423	>1 million	329–423
Size of input images	$227\times 227$	$227\times 227$	$224\times 224$	$224\times 224$
No. of output categories	1000	2	1000	2
No. of Conv. layers	5	5	16	16
FC layers	3	3	3	3
Optimizer	SGDM	SGDM	SGDM	SGDM
Loss function	cross entr.	cross entr.	cross entr.	cross entr.