COVIDNet CXR-3

[mayaliliya]

Pull Request Template

Description

Added COVIDNet CXR-3 model with MEDUSA architecture including evaluation and inference scripts.

Type of change

Please mark any boxes that apply.

• [ ] Bug fix (non-breaking change which fixes an issue)
• [X] New feature (non-breaking change which adds functionality)
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
• [ ] This change requires a documentation update

How Has This Been Tested?

Model weights were loaded from the Google Drive and then tested against the scripts. eval.py tested with:

python eval.py \
    --weightspath models/COVIDNet-CXR-3 \
    --metaname model.meta \
    --ckptname model \
    --n_classes 2 \
    --testfile labels/test_COVIDx8B.txt \
    --out_tensorname softmax/Softmax:0 \
    --is_medusa_backbone

inference.py tested with:

python inference.py \
    --weightspath models/COVIDNet-CXR-3 \
    --metaname model.meta \
    --ckptname model \
    --n_classes 2 \
    --imagepath assets/ex-covid.jpeg \
    --out_tensorname softmax/Softmax:0 \
    --is_medusa_backbone

Checklist:

Please mark any boxes that have been completed.

• [ X] I have performed a self-review of my own code.
• [ X] I have commented my code, particularly in hard-to-understand areas.
• [ X] I have made corresponding changes to the documentation.
• [ X] My changes generate no new warnings.
• [ X] Any dependent changes have been merged and published in downstream modules.

[haydengunraj]

Overall, I don't love the direction we're heading with respect to our scripts. As we add more and more models, it's becoming a bit difficult to keep track of the right tensor names, command line arguments, etc.

Ideally what we would have is a common library of all our shared code, and then a set of scripts/modules for each type of model we have. I don't think that kind of change is in the scope of this PR, but it's something to keep in mind so we don't end up with code that is cluttered and difficult to read/understand.

[mayaliliya]

Overall, I don't love the direction we're heading with respect to our scripts. As we add more and more models, it's becoming a bit difficult to keep track of the right tensor names, command line arguments, etc.

Ideally what we would have is a common library of all our shared code, and then a set of scripts/modules for each type of model we have. I don't think that kind of change is in the scope of this PR, but it's something to keep in mind so we don't end up with code that is cluttered and difficult to read/understand.

I fully agree, it is becoming difficult to add new models while attempting to make minimal changes to the pre-existing scripts. This is something we can bring up in the following meetings so we can plan a smooth refactoring process