Skip to content

pmeier/GMA_CI_2019_ssim_content_loss

Repository files navigation

This repository contains the source code to reproduce the results of the paper

"Content representation for Neural Style Transfer Algorithms based on Structural Similarity"

written by Philip Meier and Volker Lohweg. It was presented at the 29. Workshop "Computational Intelligence" on the 28th and 29th of November 2019 in Dortmund, Germany.

If you use this work within a scientific publication, please cite it as

@InProceedings{ML2019,
  author    = {Meier, Philip and Lohweg, Volker},
  title     = {Content Representation for Neural Style Transfer Algorithms based on Structural Similarity},
  booktitle = {Proceedings of the 29\textsuperscript{th} Workshop Computational Intelligence},
  year      = {2019},
  url       = {https://github.com/pmeier/GMA_CI_2019_ssim_content_loss},
}

The paper is part the conference proceedings, which are openly accessible. Unfortunately, the original paper contains several non-critical mistakes. You can find a corrected version here.

Installation

Clone this repository

git clone https://github.com/pmeier/GMA_CI_2019_ssim_content_loss

and install the required packages

cd GMA_CI_2019_ssim_content_loss
pip install -r requirements.txt

If you experience problems while installing torch or torchvision, please follow the official installation instructions for your setup.

Replication

All results are contained in the results folder. If you want to replicate the results yourself you need to

  1. download the source images by running images.py,
  2. perform the experiments by running experiments.py, and
  3. finally run process.py to process the raw experiment results.

About

Source code for the paper "Content representation for Neural Style Transfer Algorithms based on Structural Similarity"

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages