Fabian Isensee
Fabian Isensee
Just like RGB images are three channels that are always used together, nnU-Net uses your 4 modalities together. Always :-)
`pip install --upgrade setuptools` #474
Yes you will have to delete all the old training outputs if you are changing the folds. Best, Fabian
Hey there, about 1: it doesn't matter. Both work. In the current implementation nnU-Net works for n flasses (incl background) and the corner case of just one class isn't covered....
> About the deep supervision, I think 'no new U-net' should have follow the classical U-net architecture while deep supervision is not included. The experimental results disagree with this opinion...
correction: I can import `__surface_distances` with `from medpy.metric.binary import __surface_distances` but that is not good practice because `__` is telling users not to import it
+1 to that. Any updates on this?
 TX configuration, no sync on arm could be relevant!?
Thanks everyone :-)
Hi, each augmentation is applied one after the other. The only exception is SpatialTransform which applies shift, rotation, scaling and deformation at the same time for computational reasons Best, Fabian