Just curious whether winner methods would be checked on the basis of whether they are fully automated or not and producing the same test results?
Every entered approach was supposed to be fully automated, and we removed a few semi-automated approaches during review after the submission-deadline.
You may find of interest that the author of the best entry using public data in Task 3 is currently publishing the code for his highly performing approach on GitHub: GitHub - ngessert/isic2018: Code for the ISIC2018 Lesion Diagnosis Challenge
1 Like
Thanks Philipp for the update.