Unsupervised-Image-to-Image-Translation

This repository contains multiple util functions for the Unsupervised Image to Image Translation project.

Organisation

• dataset contains all the files related to the different dataset we use for the different iterations
• quantitative contains all the code to generate images for every model and compute their LPIPS score and Inception score as well as the actual generated images.
• testing contains as script to generate the commands for quick images translations
• conda_env.yml is the conda environment that we set up in order to run our ode

On the cluster

Please find the deliverables too big to be uploaded here on the cluster iccluster139.iccluster.epfl.ch:

• /ivrldata1/students/2021-spring-cs413-team3/HANDIN/INIT_dataset:
  • contains the dataset for our project:
    • cloudy, sunny, night and rainy contain what we kept from the dataset of INIT.
    • ssd contains all the classes (and images) created using RetinaNet
    • miscelaneous_and_scripts contains some scripts and intermediate txt files used during splitting.
• /ivrldata1/students/2021-spring-cs413-team3/HANDIN/models:
  • contains all the models trained that are presented in the Results part of our report as well as the baseline (FUNIT pretrained model that we fine-tuned)