What is the real difference between this model and Johnson's model?
dypromise opened this issue · comments
丁杨 commented
Dear all,
Does anybody know the real difference of the model(net) written in the code between this repository and Johnson's fast-neural-style? I mean, in this repo, the default model parameter is 'johnson', If I use the 'johnson' mode, and since the Johnson's newest version has implement 'InstanceNormalization' yet, What is the real difference? and the output is different too. Anybody knows? I'll appreciate any reply.
this repo's output:
Johnson's papers output: