Dear participants,

With one week remaining we are happy to see that there are quite some submissions being made to the development leaderboard. A word of caution however, there might be a significant performance difference between the validation and test sets. The first consists of only 50 lesions whereas the second contains 725 lesions. Given the large variety of lesions types and sizes for this task there might be a shake-up as the validation set can't fully represent the distribution of the lesions in the test set.

So keep this in mind when selecting your final submission to the test set. That model with +.005 Dice could just be overfitting to the validation set. It would be wise to keep a portion of the training data separate to test you models with.

Don't forget to prepare your test submission on time! You will only have one successful submission to the test set, but if it fails we will be here to help you fix unforeseen errors given that there is enough time before the deadline to do so... If your container works on the development leaderboard it should also work on the test set.

Kind regards,

The Challenge Organizers