search

stefanknegt / Probabilistic-Unet-Pytorch

A Probabilistic U-Net for segmentation of ambiguous images implemented in PyTorch
Apache License 2.0
270 stars 67 forks source link

data preprocessing #16

Closed ccccly closed 3 years ago

ccccly commented 3 years ago

Dear Sir, I found the model to be very sensitive to data preprocessing. When I train the model with raw LIDC data with my own preprocessing, the kl loss becomes NaN soon. Could you please describe how you do data preprocessing?

ccccly commented 3 years ago

The answer is found in Preprocessing #8