How to prepare Kitti dataset?  

Hello, 
I'd like to use semantic-segmentation with the Kitti dataset, but I cannot find any instructions. It appears that kitti.py is looking for training and testing repositories, unlike the unprocessed KITTI dataset. Is there a script for this? I would like to apply what I learn here for other image types that I convert from video. 
Thanks for your support. Best wishes, 
Michel Audette