Unsupervised-domain-adapatation-GTA-to-CITYSCAPE

1. git clone git@github.com:fredlll/Unsupervised-domain-adapatation-GTA-to-CITYSCAPE.git

	cd Unsupervised-domain-adapatation-GTA-to-CITYSCAPE
	pip -r requirements.txt 

2. Divide GTA into train, val, test parts and put in folder as follow:

/datasets
    /data
        /images                  # gta-as-cityscape images
            /1.jpg
            /2.jpg
                        ......
        /labels                 # gta-as-cityscape labels
            /1.jpg
            /2.jpg
                        ......        
        /Cityscape                
            /images                  # cityscape images
                        /1.jpg
                        /2.jpg
                        ......
            /labels                 # cityscape labels
                        /1.jpg
                        /2.jpg

3. 1st step, Train your model on GTA -as-cityscape

python main_GTAasC.py ...

4. 2nd step, Train domain adaptation using GTA -as-cityscapepretrain weight

sh train.sh