Segmentation on Test Data set

Hi all,

I am trying to test my trained model on some test data (a set of images with corresponding binary masks). My understanding is that I have to use learn.predict(test_file), where test_file is my image. However, how do I calculate my Dice and Jaccard from the prediction and my truth data (i.e. mask)? Any thoughts or pointing to an example in fast.ai v2 would be much appreciated!