search

davidiommi / Pytorch--3D-Medical-Images-Segmentation--SALMON

Segmentation deep learning ALgorithm based on MONai toolbox: single and multi-label segmentation software developed by QIMP team-Vienna.
https://zmpbmt.meduniwien.ac.at/wissenschaft-forschung/qimp/
MIT License
120 stars 26 forks source link

my train_loss is always 1.000 #4

Closed zhang5957 closed 3 years ago

davidiommi commented 3 years ago

I guess is the patch size. You want to segment something very small compared to the all patch that the network is extracting. I had the same problem, at one point the loss was shooting to 1 because the object was too small.

Otherwise, another reason is the learning rate, and you have to change it and pay attention to the optimizer as well, for that lr value.