Hi there, I trying to evaluate a gan I’ve trained, so I ended up reading about “Multi scale structural similarity”. This is also used massively in number of gan papers.
So it’d be helpful if someone in short can explain intuition behind it, and how do we calculate it.