As the annotations for “non-mitotic cells” in the data set (and also their use in the notebook) seems to have gotten people confused, let us take the chance to clarify this:
No, the non-mitotic cell annotations are not the target of the detection. They are a by-product of our annotation procedure and can help you in training your models (because they can be used as hard examples).
In the evaluation of the challenge (on the test set) they do not play any role.
Leave a Reply