When I use the pretrained model, should I use the x = (x - 0.45) / 0.225 as input?

[zhyever]

When I use the ImageNet pretrained model on other datasets, should I do something like this:

def forward(self, img):
        img = (img - 0.45) / 0.225
        x = self.mod1(img)

instead of directly :

def forward(self, img):
        x = self.mod1(img)

Need help! X_X

[zhyever]

I get it.. It is transforming the inputs for the network according to the mean and std of the KITTI dataset