Converting to BGR in style_transfer_net.py

[Mypathissional]

Hey,

I was going through the code and I cant understand why you convert the inputs from RGB to BGR.

Would be grateful for any help,

Best

[eridgd]

The original VGG19 was trained on images loaded as BGR from OpenCV and so expects that channel order. See jcjohnson/neural-style#207

The second key preprocessing step of the VGG nets is subtracting the mean pixel across ImageNet for each color channel. In this codebase that's performed here

arbitrary_style_transfer/encoder.py

Line 79 in c5730fa

return image - np.array([103.939, 116.779, 123.68])