Table 2.

CGAN model with activations units

Generative network		Discriminative network
Input	Activation	Input	Activation
Input ($1\times 1\times 100)$)	$1\times 1\times 100$	$\mathrm{Input}(64\times 64\times 3)$	$(64\times 64\times 3)$
Project and reshape	$4\times 4\times 512$	Dropout (50%)	$(64\times 64\times 3)$
Transpose convolution $5\times 5$ filter size	$8\times 8\times 256$	Convolution 1 $5\times 5$ filter size	$(32\times 32\times 64)$
Batch-normalization	$8\times 8\times 256$	LeakyReLU	$(32\times 32\times 64)$
ReLU	$8\times 8\times 256$	Convolution 2 $5\times 5$ filter size	$(16\times 16\times 128)$
Transpose convolution $5\times 5$ filter size	$16\times 16\times 128$	Batch-normalization	$(16\times 16\times 128)$
Batch-normalization	$16\times 16\times 128$	LeakyReLU	$(16\times 16\times 128)$
ReLU	$16\times 16\times 128$	Convolution 3 $5\times 5$ filter size	$(16\times 16\times 256)$
Transpose convolution $5\times 5$ filter size	$32\times 32\times 64$	Batch-normalization	$(16\times 16\times 256)$
Batch-normalization	$32\times 32\times 64$	LeakyReLU	$(8\times 8\times 128)$
ReLU	$32\times 32\times 64$	Convolution 4 $5\times 5$ filter size	$(4\times 4\times 512)$
Transpose convolution $5\times 5$ filter size	$64\times 64\times 3$	Batch-normalization	$(4\times 4\times 512)$
Hyperbolic Tanh	$64\times 64\times 3$	LeakyReLU	$(4\times 4\times 512)$
		Convolution 5 $4\times 4$ filter size	$(1\times 1\times 1)$