Table 1.

The hyper parameters of the GAN.

Operation	Kernel	Strides	Feature maps	BN	Activation
Generator
Transposed Convolution	4$\times$4	1$\times$1	512	✓	ReLU
Transposed Convolution	4$\times$4	2$\times$2	256	✓	ReLU
Transposed Convolution	4$\times$4	2$\times$2	128	✓	ReLU
Transposed Convolution	4$\times$4	2$\times$2	256	✓	ReLU
Transposed Convolution	4$\times$4	2$\times$2	3	–	Tanh
Discriminator
Convolution	4$\times$4	2$\times$2	64	–	Leaky ReLU
Convolution	4$\times$4	2$\times$2	128	✓	Leaky ReLU
Convolution	4$\times$4	2$\times$2	256	✓	Leaky ReLU
Convolution	4$\times$4	2$\times$2	512	✓	Leaky ReLU
Convolution	4$\times$4	1$\times$1	1	–	Sigmoid
Frame Work	PyTorch
Batch size for real data	128
Batch size for generator	128
Number of worker	2
Number of epochs	400
Optimizer	Adam (${\beta }_1=0.5$)
Learning rate	0.0002
Leaky ReLU slope	0.2
Loss function	Binary Cross Entropy