Table 2.

Hyperparameters used to train the CNNs used in our experiments.

Hyperparameter	Value
Input dimension	100 × 100
Number of convolution layers	2
Number of fully connected layers	1
Number of filters for each convolution layer	32, 64
Size of convolutional kernels	3 × 3
Strides size	2
Activation function for hidden layers	ReLU
Loss function	Hinge
L2 regularization coefficient	0.001
Number of neurons of fully connected layers	128
Batch size	256