Don't get your hopes up just yet. It's not usable in its current state, unless you have Matlab (octave doesn't work, see closed+open Github issues). There are efforts under way to port the matlab code to python, but i haven't tried the fork yet. Also, processing a single image on a K80 takes several hours, no comparison to neural style. Likewise, generating the laplacians also takes a loooong time, at least with octave. Maybe this will be accelerated in an python version.
