• https://github.com/Shanghai-Aitrox-Technology/EfficientSegmentation/blob/master/Common/image_resample.py
• https://github.com/xyndameinv/FLARE21/blob/main/process1.py

• Try different losses (default is STCN's OhemCE)
• #25
• Choose reference based on organs's sizes or adapt active learning. Experiment other different referencing strategies.

Double check why is unit testing failed?

https://docs.monai.io/en/latest/visualize.html

Check file not found/folder exist/path validate before running inference script

In README.md, there is a line mentioned about requirements.txt:

Go to root folder, run pip install -r requirements.txt

However, there is no such file in root folder, instead the folder contains environment.yml. I think there is a mistake in document and you should fix it ^^.

• Optimize propagate strategy with low resource
• Design new pattern of multiple strategy
• Time computing for each module / part of strategy
• Add unit test

• While training, use intermediate slice (which must have all classes) to be the guiding frame.
• Try cropping out black values from tensors
• Look for more preprocessing, postprocesing functions from nnUnet which can be adapted with ease
• Instead of batchifying 3 frames from different volumes, is sampling batches within a volume better?

• Proposal:

  • Add UE to loss calculation
  • Calculate variance between two models from CPS, then select which pseudo labels to be used (binary)

• https://github.com/layumi/Seg-Uncertainty

• https://arxiv.org/abs/1706.04737

• https://arxiv.org/pdf/2003.03773v3.pdf