@etetteh since you are 'intern' to the core dev team, I suggest you look at the code that generates this report to figure out what it does, and if needed update the tutorial and/or output display to make it more intuitive to future users.

Perhaps needless to say, but a comprehensive understanding of the

function and stepping inside the sub-functions within (e.g. imed_evaluation for evaluation) will go a long way!

Am I right to say that, what is happening there is, random subjects are selected, and the model does a prediction on them?

Yes, you are right. Test subjects are selected based on the "random_seed" key in the config file (described here). Using the same seed across different runs (e.g. ivadomed --train and then ivadomed --test) ensures the same set of subjects to be selected as test.