Write project description
Open Source, Open Science project. Building a strong semi-supervised model that can be used for any CT Lung based task.
Data Normalization strategy
Should we use the lung window?
Should we normalize the data before saving?
What is the format used in the datasets we will use?
Upload to Dataverse
Dataverse onboarding
- [ ] Download datasets we will use
- [ ] Normalize data
- [ ] Update to have a consistent format
- If we want to share the data publicly, we have to carefully think about the license
Design candidate model architectures
Check ****Models section.
- Design the pipeline for semi-supervised training(SST) of the base model
- Design the set of augmentation(candidates) used for SST
Supervised training\scoring pipeline
For evaluating the base model we will use it as a starting point for supervised model and will log best accuracy after N epochs or after early stopping.