Pull Request Template

Description

This adds a script which performs transfer-learning on an existing COVIDNet to produce COVIDNet-Risk a risk-stratification network that predicts patient offset (days since symptom onset).

Eventually this may be used to help make more informed decisions about stage of care, but currently there is not enough data to demonstrate complete examples (52 train 9 test).

There is also work to be done - which is left in the form of TODOs and FIXMES. we can migrate these to issues if we like.

Context of change

[x] Software (software that runs on the PC)
[ ] Library (library that runs on the PC)
[ ] Tool (tool that assists coding development)
[ ] Other

Type of change

[x] Bug fix (non-breaking change which fixes an issue)
[ ] 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?

I've tested this by running the script locally, it trains well on my laptop (mobile GTX1080)a by default it produces <CovidNet>/models/COVIDNet-Risk/COVIDNet-Risk-lr2e-06

Checklist:

[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.

example output:

Saved pre-trained model with re-initialized output layers.
Baseline eval:
    evaluated 9 images.
confusion matrix:
[[0. 0. 0. 0.]
 [0. 0. 0. 0.]
 [1. 1. 0. 1.]
 [6. 0. 0. 0.]]
per-class accuracies:
[0, 0, 0.0, 0.0]
Transfer Learning Started.
    train samples: 52
    test samples:  9
    stratification: [3, 5, 10]

Fine-Tuning on 1 epoch = 52 images.
...

lindawangg / COVID-Net