|
|
@ -214,35 +214,41 @@ class GTSRBmodel(nn.Module): |
|
|
|
|
|
|
|
self.feature = nn.Sequential( |
|
|
|
nn.Conv2d(in_channels=3, out_channels=96, kernel_size=(5, 5)), # 36 ; 44 |
|
|
|
nn.BatchNorm2d(96), |
|
|
|
nn.ReLU(inplace=True), |
|
|
|
nn.MaxPool2d(kernel_size=(2, 2), stride=(2, 2)), # 18 ; 22 |
|
|
|
nn.Conv2d(in_channels=96, out_channels=144, kernel_size=(3, 3)), # 16 ; 20 |
|
|
|
nn.BatchNorm2d(144), |
|
|
|
nn.ReLU(inplace=True), |
|
|
|
nn.MaxPool2d(kernel_size=(2, 2), stride=(2, 2)), # 8 ; 10 |
|
|
|
nn.Conv2d(in_channels=144, out_channels=256, kernel_size=(5, 5)), # 4 ; 6 |
|
|
|
nn.BatchNorm2d(256), |
|
|
|
nn.Dropout2d(), |
|
|
|
nn.ReLU(inplace=True), |
|
|
|
nn.MaxPool2d(kernel_size=(2, 2), stride=(2, 2)), # 2 ; 3 |
|
|
|
) |
|
|
|
|
|
|
|
self.classifier = nn.Sequential( |
|
|
|
nn.Linear(256 * 3 * 3, 512), |
|
|
|
nn.Linear(256 * 2 * 2, 512), |
|
|
|
nn.BatchNorm1d(512), |
|
|
|
nn.ReLU(inplace=True), |
|
|
|
nn.Linear(512, 43), |
|
|
|
) |
|
|
|
|
|
|
|
self.discriminator = nn.Sequential( |
|
|
|
nn.Linear(256 * 3 * 3, 1024), |
|
|
|
nn.Linear(256 * 2 * 2, 1024), |
|
|
|
nn.BatchNorm1d(1024), |
|
|
|
nn.ReLU(inplace=True), |
|
|
|
nn.Linear(1024, 1024), |
|
|
|
nn.BatchNorm1d(1024), |
|
|
|
nn.ReLU(inplace=True), |
|
|
|
nn.Linear(1024, 2), |
|
|
|
) |
|
|
|
|
|
|
|
def forward(self, input_data, alpha = 1.0): |
|
|
|
input_data = input_data.expand(input_data.data.shape[0], 3, 48, 48) |
|
|
|
input_data = input_data.expand(input_data.data.shape[0], 3, 40, 40) |
|
|
|
feature = self.feature(input_data) |
|
|
|
feature = feature.view(-1, 256 * 3 * 3) |
|
|
|
feature = feature.view(-1, 256 * 2 * 2) |
|
|
|
reverse_feature = ReverseLayerF.apply(feature, alpha) |
|
|
|
class_output = self.classifier(feature) |
|
|
|
domain_output = self.discriminator(reverse_feature) |
|
|
|