direct icon indicating copy to clipboard operation
direct copied to clipboard

New loss functions, refactored engines to only implement forward_method, quality fixes

Open georgeyiasemis opened this issue 1 year ago • 1 comments

Implements:

  • New loss functions (NMSE, NRMSE, NMAE, SobelGradL1Loss, SobelGradL2Loss)
  • mri_models performs _do_iteration method, child engines perform forward_function which returns output_image and/or output_kspace
  • Changes/Additions in mri_transforms
    • Padding computed as a Tensor with ComputePadding transform (this is helpful when cropping image and tranforming to kspace)
    • ApplyPadding transform
    • ComputeImage transform choices of mod output or not
    • RenameKeys transform
    • Normalize split to ComputeScalingFactor and Normalize
  • Some quality changes
  • Some documentation changes

georgeyiasemis avatar Oct 17 '22 13:10 georgeyiasemis

Codecov Report

Base: 82.02% // Head: 81.91% // Decreases project coverage by -0.10% :warning:

Coverage data is based on head (09e1ba6) compared to base (1a8ddf5). Patch coverage: 92.57% of modified lines in pull request are covered.

:exclamation: Current head 09e1ba6 differs from pull request most recent head 8220124. Consider uploading reports for the commit 8220124 to get more accurate results

Additional details and impacted files
@@            Coverage Diff             @@
##             main     #224      +/-   ##
==========================================
- Coverage   82.02%   81.91%   -0.11%     
==========================================
  Files          74       77       +3     
  Lines        5585     5590       +5     
==========================================
- Hits         4581     4579       -2     
- Misses       1004     1011       +7     
Impacted Files Coverage Δ
direct/nn/cirim/cirim.py 98.27% <ø> (ø)
direct/nn/cirim/cirim_engine.py 95.45% <50.00%> (ø)
direct/exceptions.py 62.50% <85.71%> (+21.32%) :arrow_up:
direct/nn/mri_models.py 91.30% <85.98%> (-2.31%) :arrow_down:
direct/data/mri_transforms.py 97.00% <89.47%> (-2.29%) :arrow_down:
direct/functionals/grad.py 97.95% <97.95%> (ø)
direct/common/subsample.py 84.63% <100.00%> (ø)
direct/data/datasets.py 87.46% <100.00%> (-0.99%) :arrow_down:
direct/data/datasets_config.py 100.00% <100.00%> (ø)
direct/data/h5_data.py 65.06% <100.00%> (+0.24%) :arrow_up:
... and 20 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.

codecov[bot] avatar Oct 17 '22 13:10 codecov[bot]