stylegan2-pytorch
stylegan2-pytorch copied to clipboard
ability to calculate Perpectual Path Length (PPL)?
Hello,
In the original paper they use Perceptual Path Length (PPL) to evaluate quality of models that have identical FID. Is there an implementation of PPL that can be used (akin to how FID can be added)?
I recognize that PPL is basically integrated into the model through the path length regularization that occurs periodically throughout training, but I'd like to be able to calculate PPL directly as well.
Thanks!
I've found this tutorial on calculating PPL...