The decoder network in the generator network also consists of eight upsampling convolutional blocks. The configuration for the eight upsampling convolutional blocks is as follows:
|
Layer Name |
Hyperparameters |
Input Shape |
Output Shape |
|
1st 2D Upsampling Layer |
size=(2, 2) |
(1, 1, 512) |
(2, 2, 512) |
|
2D Convolution Layer |
filters=512, kernel_size=4, strides=1, padding='same', |
(2, 2, 512) |
(2, 2, 512) |
|
Batch Normalization Layer |
None |
(2, 2, 512) |
(2, 2, 512) |
|
Dropout Layer |
dropout=0.5 |
(2, 2, 512) |
(2, 2, 512) |
|
Concatenation Layer (7th Conv layer from the encoder network) |
axis=3 |
(2, 2, 512) |
(2, 2, 1024) |
|
Activation Layer |
activation='relu' |
(2, 2, 1024) |
(2, 2, 1024) |
