Open ogencoglu opened 3 months ago
Hi @ogencoglu,
thanks for your interest in Skeleton Recall! We will add the evaluation script we used soonish, it just needs some cleaning up ;). We didn't perform early stopping for the nnUNet based trainings but used the default training schedule and took the final checkpoint.
Best, Yannick
Thanks for the swift reply! Will be waiting.
Thanks for the great work!
Do you have the evaluation (especially Betti errors) script somewhere for the 2D binary segmentation task?
Another question is did you perform early stopping, learning rate scheduling etc based on Skeleton Recall loss or on the final metrics you reported, i.e. clDice and Betti stuff?