Dear Felix,
Any rotation of the paper ECGs in the images is possible, but most are unlikely. The hidden validation and test sets are intended to be realistic, not adversarial or antagonistic. :)
I would suggest thinking about the kinds of distortions that are likely to happen when someone photographs or scans paper ECGs and focusing on them. Of course, a model may not perform as well on certain rare distortions, but, ideally, it should also not crash if it encounters them, too.
Best,
Matt
(On behalf of the Challenge team.)
Please post questions and comments in the forum. However, if your question reveals information about your entry, then please email info at
physionetchallenge.org. We may post parts of our reply publicly if we feel that all Challengers should benefit from it. We will not answer emails about the Challenge to any other address. This email is maintained by a group. Please do not email us individually.