def __init__(self, init_weights_path=None):
super(Discriminator, self).__init__()
# Activation Functions
self.LeakyReLU = nn.LeakyReLU(0.2, True)
self.Sigmoid = nn.Sigmoid()
## 1st Block
self.conv_01_1 = nn.Conv2d(3, 32, 3, 1, 1)
# LeakyReLU
## 2nd Block
self.conv_02_1 = nn.Conv2d(32, 64, 3, 2, 1)
# LeakyReLU
self.conv_02_2 = nn.Conv2d(64, 128, 3, 1, 1)
self.in_norm_02_1 = InstanceNormalization(128)
# LeakyReLU
## 3rd Block
self.conv_03_1 = nn.Conv2d(128, 128, 3, 2, 1)
# Leaky ReLU
self.conv_03_2 = nn.Conv2d(128, 256, 3, 1, 1)
self.in_norm_03_1 = InstanceNormalization(256)
# LeakyReLU
## 4th Block
self.conv_04_1 = nn.Conv2d(256, 256, 3, 1, 1)
self.in_norm_04_1 = InstanceNormalization(256)
# LeakyReLU
## Final Block
self.conv_05_1 = nn.Conv2d(256, 1, 3, 1, 1)
# Sigmoid
# Initialize weights
if init_weights_path is None:
init_weights(self)
else:
self.load_state_dict(
torch.load(init_weights_path)["model_state_dict"])
def __init__(self, init_weights_path=None):
super(Inception_v3, self).__init__()
self.Conv2d_1a_3x3 = BasicConv2d(3, 32, kernel_size=3, stride=2)
self.Conv2d_2a_3x3 = BasicConv2d(32, 32, kernel_size=3)
self.Conv2d_2b_3x3 = BasicConv2d(32, 64, kernel_size=3, padding=1)
self.maxpool1 = nn.MaxPool2d(kernel_size=3, stride=2)
self.Conv2d_3b_1x1 = BasicConv2d(64, 80, kernel_size=1)
self.Conv2d_4a_3x3 = BasicConv2d(80, 192, kernel_size=3)
self.maxpool2 = nn.MaxPool2d(kernel_size=3, stride=2)
self.Mixed_5b = InceptionA(192, pool_features=32)
self.Mixed_5c = InceptionA(256, pool_features=64)
self.Mixed_5d = InceptionA(288, pool_features=64)
self.Mixed_6a = InceptionB(288)
self.Mixed_6b = InceptionC(768, channels_7x7=128)
self.Mixed_6c = InceptionC(768, channels_7x7=160)
self.Mixed_6d = InceptionC(768, channels_7x7=160)
self.Mixed_6e = InceptionC(768, channels_7x7=192)
# Initialize weights
if init_weights_path is None:
init_weights(self)
else:
self.load_state_dict(torch.load(init_weights_path))
def __init__(self, init_weights_path=None):
super(Generator, self).__init__()
# Activation Functions
self.ReLU = nn.ReLU()
self.Tanh = nn.Tanh()
## First Block
self.pad_01_1 = nn.ReflectionPad2d(3)
self.conv_01_1 = nn.Conv2d(3, 64, 7)
self.in_norm_01_1 = InstanceNormalization(64)
# ReLU
## Down Convolution Block 1
self.conv_02_1 = nn.Conv2d(64, 128, 3, 2, 1)
self.conv_02_2 = nn.Conv2d(128, 128, 3, 1, 1)
self.in_norm_02_1 = InstanceNormalization(128)
# ReLU
## Down Convolution Block 2
self.conv_03_1 = nn.Conv2d(128, 256, 3, 2, 1)
self.conv_03_2 = nn.Conv2d(256, 256, 3, 1, 1)
self.in_norm_03_1 = InstanceNormalization(256)
# ReLU
## Residual Block 1
self.pad_04_1 = nn.ReflectionPad2d(1)
self.conv_04_1 = nn.Conv2d(256, 256, 3)
self.in_norm_04_1 = InstanceNormalization(256)
# ReLU
self.pad_04_2 = nn.ReflectionPad2d(1)
self.conv_04_2 = nn.Conv2d(256, 256, 3)
self.in_norm_04_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 2
self.pad_05_1 = nn.ReflectionPad2d(1)
self.conv_05_1 = nn.Conv2d(256, 256, 3)
self.in_norm_05_1 = InstanceNormalization(256)
# ReLU
self.pad_05_2 = nn.ReflectionPad2d(1)
self.conv_05_2 = nn.Conv2d(256, 256, 3)
self.in_norm_05_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 3
self.pad_06_1 = nn.ReflectionPad2d(1)
self.conv_06_1 = nn.Conv2d(256, 256, 3)
self.in_norm_06_1 = InstanceNormalization(256)
# ReLU
self.pad_06_2 = nn.ReflectionPad2d(1)
self.conv_06_2 = nn.Conv2d(256, 256, 3)
self.in_norm_06_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 4
self.pad_07_1 = nn.ReflectionPad2d(1)
self.conv_07_1 = nn.Conv2d(256, 256, 3)
self.in_norm_07_1 = InstanceNormalization(256)
# ReLU
self.pad_07_2 = nn.ReflectionPad2d(1)
self.conv_07_2 = nn.Conv2d(256, 256, 3)
self.in_norm_07_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 5
self.pad_08_1 = nn.ReflectionPad2d(1)
self.conv_08_1 = nn.Conv2d(256, 256, 3)
self.in_norm_08_1 = InstanceNormalization(256)
# ReLU
self.pad_08_2 = nn.ReflectionPad2d(1)
self.conv_08_2 = nn.Conv2d(256, 256, 3)
self.in_norm_08_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 6
self.pad_09_1 = nn.ReflectionPad2d(1)
self.conv_09_1 = nn.Conv2d(256, 256, 3)
self.in_norm_09_1 = InstanceNormalization(256)
# ReLU
self.pad_09_2 = nn.ReflectionPad2d(1)
self.conv_09_2 = nn.Conv2d(256, 256, 3)
self.in_norm_09_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 7
self.pad_10_1 = nn.ReflectionPad2d(1)
self.conv_10_1 = nn.Conv2d(256, 256, 3)
self.in_norm_10_1 = InstanceNormalization(256)
# ReLU
self.pad_10_2 = nn.ReflectionPad2d(1)
self.conv_10_2 = nn.Conv2d(256, 256, 3)
self.in_norm_10_2 = InstanceNormalization(256)
# Elementwise Sum
## Residual Block 8
self.pad_11_1 = nn.ReflectionPad2d(1)
self.conv_11_1 = nn.Conv2d(256, 256, 3)
self.in_norm_11_1 = InstanceNormalization(256)
# ReLU
self.pad_11_2 = nn.ReflectionPad2d(1)
self.conv_11_2 = nn.Conv2d(256, 256, 3)
self.in_norm_11_2 = InstanceNormalization(256)
# Elementwise Sum
## Up Convolution Block 1
self.deconv_12_1 = nn.ConvTranspose2d(256, 128, 3, 2, 1, 1)
self.deconv_12_2 = nn.Conv2d(128, 128, 3, 1, 1)
self.in_norm_12_1 = InstanceNormalization(128)
# ReLU
## Up Convolution Block 2
self.deconv_13_1 = nn.ConvTranspose2d(128, 64, 3, 2, 1, 1)
self.deconv_13_2 = nn.Conv2d(64, 64, 3, 1, 1)
self.in_norm_13_1 = InstanceNormalization(64)
# ReLU
## Final Block
self.pad_14_1 = nn.ReflectionPad2d(3)
self.conv_14_1 = nn.Conv2d(64, 3, 7)
# tanh
# Initialize weights
if init_weights_path is None:
init_weights(self)
else:
self.load_state_dict(
torch.load(init_weights_path)["model_state_dict"])
def __init__(self, init_weights_path=None):
super(VGG19, self).__init__()
self.features = nn.Sequential(
# Block 1
nn.Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(64,
64,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=2,
stride=2,
padding=0,
dilation=1,
ceil_mode=False),
# Block 2
nn.Conv2d(64,
128,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(128,
128,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=2,
stride=2,
padding=0,
dilation=1,
ceil_mode=False),
# Block 3
nn.Conv2d(128,
256,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(256,
256,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(256,
256,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(256,
256,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=2,
stride=2,
padding=0,
dilation=1,
ceil_mode=False),
# Block 4
nn.Conv2d(256,
512,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(512,
512,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(512,
512,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
nn.Conv2d(512,
512,
kernel_size=(3, 3),
stride=(1, 1),
padding=(1, 1)),
nn.ReLU(inplace=True),
)
# Initialize weights
if init_weights_path is None:
init_weights(self)
else:
self.load_state_dict(
torch.load(init_weights_path)["model_state_dict"])
