SageMaker pipeline

[joeranbosma]

Update script to prepare detection dataset:

• Perform inference with classification model
• Perform inference with segmentation model
• Prepare detection dataset

Also:

• Allow setting paths in preprocessing functions
• Bugfix (hardcoded 1500 as number of cases)
• Rename training scripts to step_n_...

[joeranbosma]

The proposed changes result in code that does not work, it fails with the following error:

Traceback (most recent call last):
  File "/opt/algorithm/ITUNet-for-PICAI-2022-Challenge/step_3_prepare_detection_data.py", line 73, in <module>
    main()
  File "/opt/algorithm/ITUNet-for-PICAI-2022-Challenge/step_3_prepare_detection_data.py", line 36, in main
    predict_test5c(
  File "/opt/algorithm/ITUNet-for-PICAI-2022-Challenge/classification/cls_data.py", line 133, in predict_test5c
    weight_list = get_weight_list(weight_path,choice=[1,2,3,4,5])
  File "/opt/algorithm/ITUNet-for-PICAI-2022-Challenge/classification/cls_data.py", line 46, in get_weight_list
    if choice is not None and eval(str(fold.name)[-1]) not in choice:
  File "<string>", line 1, in <module>
NameError: name 'h' is not defined

The issue seems to be in eval(str(fold.name)[-1]).

[joeranbosma]

When testing the code some more issues came up. I've now fixed these issues. Please review the code and merge it in if you agree with the changes.

[joeranbosma]

I've now also added the code for the final training step: training the semi-supervised segmentation model.