We welcome pull requests to improve functionality of the library. Please discuss with the team before making a pull request to make sure it fits the scope of the project.
Here are some directions that we would like to go in:
- Tuberculosis datasets.
- Segmentations for existing datasets (pathology masks, organ segmentations, object segmentations)
- object-CXR https://jfhealthcare.github.io/object-CXR/
- Models from the CheXpert leaderboard https://stanfordmlgroup.github.io/competitions/chexpert/
One issue here is that they are interchangable with the existing model and are well organized.
Evaluation scripts?