models/networks/architectures.py [212:250]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def forward(self, input):
        # Encoder
        # Convolution layers:
        # input is (nc) x 256 x 256
        e1 = self.conv1(input)
        # state size is (num_filters) x 128 x 128
        e2 = self.batch_norm2_0(self.conv2(self.leaky_relu(e1)))
        # state size is (num_filters x 2) x 64 x 64
        e3 = self.batch_norm4_0(self.conv3(self.leaky_relu(e2)))
        # state size is (num_filters x 4) x 32 x 32
        e4 = self.batch_norm8_0(self.conv4(self.leaky_relu(e3)))
        # state size is (num_filters x 8) x 16 x 16
        e5 = self.batch_norm8_1(self.conv5(self.leaky_relu(e4)))
        # state size is (num_filters x 8) x 8 x 8
        e6 = self.batch_norm8_2(self.conv6(self.leaky_relu(e5)))
        # state size is (num_filters x 8) x 4 x 4
        e7 = self.batch_norm8_3(self.conv7(self.leaky_relu(e6)))
        # state size is (num_filters x 8) x 2 x 2
        # No batch norm on output of Encoder
        e8 = self.conv8(self.leaky_relu(e7))

        # Decoder
        # Deconvolution layers:
        # state size is (num_filters x 8) x 1 x 1
        d1_ = self.batch_norm8_4(self.dconv1(self.up(self.relu(e8))))
        # state size is (num_filters x 8) x 2 x 2
        d1 = torch.cat((d1_, e7), 1)
        d2_ = self.batch_norm8_5(self.dconv2(self.up(self.relu(d1))))
        # state size is (num_filters x 8) x 4 x 4
        d2 = torch.cat((d2_, e6), 1)
        d3_ = self.batch_norm8_6(self.dconv3(self.up(self.relu(d2))))
        # state size is (num_filters x 8) x 8 x 8
        d3 = torch.cat((d3_, e5), 1)
        d4_ = self.batch_norm8_7(self.dconv4(self.up(self.relu(d3))))
        # state size is (num_filters x 8) x 16 x 16
        d4 = torch.cat((d4_, e4), 1)
        d5_ = self.batch_norm4_1(self.dconv5(self.up(self.relu(d4))))
        # state size is (num_filters x 4) x 32 x 32
        d5 = torch.cat((d5_, e3), 1)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



models/networks/architectures.py [394:432]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def forward(self, input):
        # Encoder
        # Convolution layers:
        # input is (nc) x 256 x 256
        e1 = self.conv1(input)
        # state size is (num_filters) x 128 x 128
        e2 = self.batch_norm2_0(self.conv2(self.leaky_relu(e1)))
        # state size is (num_filters x 2) x 64 x 64
        e3 = self.batch_norm4_0(self.conv3(self.leaky_relu(e2)))
        # state size is (num_filters x 4) x 32 x 32
        e4 = self.batch_norm8_0(self.conv4(self.leaky_relu(e3)))
        # state size is (num_filters x 8) x 16 x 16
        e5 = self.batch_norm8_1(self.conv5(self.leaky_relu(e4)))
        # state size is (num_filters x 8) x 8 x 8
        e6 = self.batch_norm8_2(self.conv6(self.leaky_relu(e5)))
        # state size is (num_filters x 8) x 4 x 4
        e7 = self.batch_norm8_3(self.conv7(self.leaky_relu(e6)))
        # state size is (num_filters x 8) x 2 x 2
        # No batch norm on output of Encoder
        e8 = self.conv8(self.leaky_relu(e7))

        # Decoder
        # Deconvolution layers:
        # state size is (num_filters x 8) x 1 x 1
        d1_ = self.batch_norm8_4(self.dconv1(self.up(self.relu(e8))))
        # state size is (num_filters x 8) x 2 x 2
        d1 = torch.cat((d1_, e7), 1)
        d2_ = self.batch_norm8_5(self.dconv2(self.up(self.relu(d1))))
        # state size is (num_filters x 8) x 4 x 4
        d2 = torch.cat((d2_, e6), 1)
        d3_ = self.batch_norm8_6(self.dconv3(self.up(self.relu(d2))))
        # state size is (num_filters x 8) x 8 x 8
        d3 = torch.cat((d3_, e5), 1)
        d4_ = self.batch_norm8_7(self.dconv4(self.up(self.relu(d3))))
        # state size is (num_filters x 8) x 16 x 16
        d4 = torch.cat((d4_, e4), 1)
        d5_ = self.batch_norm4_1(self.dconv5(self.up(self.relu(d4))))
        # state size is (num_filters x 4) x 32 x 32
        d5 = torch.cat((d5_, e3), 1)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -