def forward(self, x):
batch_size = x.size(0)
net = normalize_imagenet(x)
net = self.convnet(net)
net = net.view(batch_size, 256*3*3)
out = self.fc_out(net)
return out
def forward(self, x):
if self.normalize:
x = normalize_imagenet(x)
x0, x1, x2 = self.features(x)
gray_x = gray(x)
gaussian1 = self.gaussion_conv1(gray_x)
gaussian2 = self.gaussion_conv2(gray_x)
gaussian3 = self.gaussion_conv3(gray_x)
gaussian4 = self.gaussion_conv4(gray_x)
gaussian5 = self.gaussion_conv5(gray_x)
gaussian6 = self.gaussion_conv6(gray_x)
dog1 = torch.sub(gaussian2, gaussian1)
dog2 = torch.sub(gaussian4, gaussian3)
dog3 = torch.sub(gaussian6, gaussian5)
dog_tem = torch.cat((dog1, dog2), dim=1)
dog = torch.cat((dog_tem, dog3), dim=1)
out_dog = self.dog_encoder(dog)
out0 = self.fc(x0)
out1 = self.fc_head1(x1)
out2 = self.fc_head2(x2)
out_dog = self.fusion_dog_ori(torch.cat((out_dog, out0), dim=1))
attention_out_dog = self.attention(out_dog.unsqueeze(dim=-1))
return out0, out1, out2, attention_out_dog.squeeze(1)
def forward(self, x):
if self.normalize:
x = normalize_imagenet(x)
net = self.features(x)
out = self.fc(net)
return out