I couldn’t find the definition of per-pixel loss and am struggling to understand the difference between per-pixel loss and content(perceptual) loss. I understand that
both loss functions measure how similar 2 images are alike.
for content loss you can combine multiple layers so that’s an advantage.
In terms of measuring the loss, are they both measuring the per-pixel differences? if not, how are they measuring things differently? thank you!
Can you verify if the following description is correct? If not, can you point out the errors for me?
for “per-pixel loss”, we are summing all the absolute error between all pixels.
for content loss, we are summing all the squared errors between all pixels and take the mean.