Thanks, I will look into that. It's not altogether clear from their site — their Docker image basically implies that you're supposed to use their pre-existing datasets, which aren't free.