Wrong normalization?

[lars76]

Where does the mean/std come from in ScalingLayer? The VGG16 mean/std is mean = [0.485, 0.456, 0.406] and std = [0.229, 0.224, 0.225].

[richzhang]

Thanks for the question. This has caused some confusion. The mean/std is recomputed to expect images from [-1, +1] instead of [0, +1] (which is what the values are for above). After the scaling layer, the results are the same.

[lars76]

Thank you for the quick response. I should have read more carefully the other issues :)

Here the computation in detail if somebody comes across this issue. Let us say we have a pixel in the R-channel with a pixel value of 240. I rescale it to [-1, 1] with 2 * 240/255 - 1 = 15/17. Then I normalize it with ScalingLayer and get (15/17 - -0.030)/0.458 = 1.99203. Regular VGG would do (240/255 - 0.485)/0.229 = 1.99203. So the normalization is indeed correct.

[98mxr]

I have a small problem. Since it is to be compatible with imagenet's mean/std, it is not good to input the img of [0, 1] from the beginning.

[Zyriix]

Still confused about this.
If my inputs is already normalized by imagenet mean std.
Eg: x (0-1)
x = (x - imgnet_mean)/imgnet_val

if i do this before i put it into lpips , does it means that i can just delete the scaling layers?

[richzhang]

Yes, that's right

[Zyriix]

Yes, that's right

ok, thank you : )