COVID-Net Open Source Initiative - COVIDNet-CT

https://github.com/haydengunraj/COVIDNet-CT

This looks like a perfect starting point. The repository contains models, training, validation, augmentation scripts for CT data. It contains steps and scripts for building one of the biggest public pneumonia CT datasets available: https://github.com/haydengunraj/COVIDNet-CT/blob/master/docs/dataset.md

NOTE: we can combine the dataset with MosMedData, ieee8023/covid-chestxray-dataset, kaggle/osic-pulmonary-fibrosis-progression, UCSD-AI4H/COVID-CT

EfficientNet-C19: Towards an Effective and Efficient Deep Learning Model for COVID-19 Patterns Detection in X-ray Images

https://github.com/ufopcsilab/EfficientNet-C19

They report a very high performance of the model. The model is built for X-ray images but we will build a 2d model as well, so there should be no problem applying the same methods for our model. We can check if we find that the default model is computationally too expensive. Then we can use teacher-student approach to train an EfficientNet.

A Weakly-supervised Framework for COVID-19 Classification and Lesion Localization from Chest CT

https://github.com/sydney0zq/covid-19-detection

TODO: check their weakly-supervised approach. Is there something we can use for our models?

COVID-19 Detection Neural Network (COVNet)

https://github.com/bkong999/COVNet

We can check it as an example architecture and look for any usefull info in their research.