code-repository

Getting started

Create a local clone using command:

git clone https://github.com/dose-iitd/code-repository

Use environment.yml file provided to reproduce the conda environment using

conda env create -f environment.yml

Download the datasets and change location cs_path, nd_path, dz_path, fd_path, fz_path, acdc_path in train.py file for Cityscapes, Nighttime driving, Dark Zurich, Foggy Driving, Foggy Zurich and ACDC datasets respectively. Pre-trained model path are passed as command line arguments in shell script files provided.

All pre-trained models can be downloaded using the following [Download link]

Pretrained models used: IBNNet, Switchable Whitening, RobustNet

sh train_dg.sh