Grand Challenge Metrics contains useful functions for evaluating machine learning models in the context of medical imaging.
- Free software: MIT license
- Documentation: https://diagnijmegen.github.io/rse-grand-challenge-metrics/.
- Bounding box annotations with Jaccard Index calculations
- Calculations of bootstrapped ROC curves
- Scoring for detection tasks
- Efficient calculation of confusion matrices, jaccard scores, dice scores, hausdorff distances, absolute volume differences, and relative volume differences
Grand Challenge Metrics requires Python 3.10 or above, and can be installed from pip.
Please see the Getting Started documentation for more details.