search

rrmina / fast-neural-style-pytorch

Fast Neural Style Transfer implementation in PyTorch :art: :art: :art:
321 stars 77 forks source link

Difference between this implementation and PyTorch team's version? #22

Open chrischen opened 2 years ago

chrischen commented 2 years ago

What is the difference of this implementation (regarding just static image style transfer, not for videos) vs the PyTorch teams implementation: the PyTorch TeamPyTorch Examples: fast-neural-style?

rrmina commented 2 years ago

They are basically the same except for 2 major things:

  1. This repo uses the original VGG networks used in the original fast-neural-style paper. In contrast, the Pytorch Team example uses VGG networks from Pytorch Image Model Zoo.
  2. This repo uses (1) Content Loss, (2) Style Loss, and (3) Total Variation Loss, while Pytorch Team example uses only (1) and (2). Total Variation Loss smooths the pixels of the output image by minimizing the difference of pixel values of adjacent pixels.

TL;DR This repo is a more faithful reproduction of the original fast-neural-style paper, but both this and Pytorch's should more or less have similar results! :)))