The code and resouces for SSF-DAN , We have uploaded all the models and the evaluate code.
We will publish the paper and upload the train code soon.

System : ubuntu 16.04
Hardware: Nvidia Tesla P100
Software: Pytorch 0.4.1, CUDA-8.0, cuDNN 5.1 or 6.0, python 3.5
For more packages information, we prepared them in the requires.sh file.

We released our models in google drive, they can be downloaded [here]

In the experiments, we use GTA5, cityscape, Synthia, Crosscity datasets,you can download them in the follow links:
[GTA5] [Cityscape] [Synthia] [Crosscity]

1.make a project dir:

mkdir ~/SSF-DAN

2.clone the SSF-DAN repository:

git clone https://github.com/JingangTan/SSF-DAN

3.download the GTA5 dataset and put it in dataset/GTA5/ and download Cityscape dataset put it in dataset/Cityscapes/. For other dataset, you can take the same operation.

4.create a new virtual enviromen, if you want to use system python enviromen,you can ignore this.

sudo pip install virtualenv
virtualenv -p /usr/bin/python3.5 ~/vir-ssf-dan
cd ~/vir-ssf-dan
source bin/activate
cd ~/SSF-DAN

5.install the packages in requires.sh:

sh requires.sh

python evaluate.py
iou.py