When test, Why normalize using the whole test dataset's features?

[huyangc]

I think the codes here shows that you normalize each feature on the whole test dataset's features.

It is some tricks or I misunderstood the codes?

[nttstar]

This is just for calculating average L2 norm of all test feature embeddings.

[huyangc]

thank you! And I have another question about the input image. As the code here shows that the only thing to do is to input the image cropped by detection bounding box, am I right? No need to enlarge the bounding box? Because some tight box only contains the face.

And thank you very much to implement so many loss functions in mxnet, because I am a big fan of mxnet.

[nttstar]

some margin is preferred.

[huyangc]

Thanks!