Training Experiment · Python
Do multimodal skin lesion classifiers lean on metadata shortcuts? Image vs metadata vs fusion on ISIC 2019, with occlusion, subgroup and fairness tests.
The repository investigates shortcut learning in multimodal skin lesion classification by training and evaluating image-only, metadata-only, and fusion models, and running experiments to assess reliance on demographic shortcuts. It includes data preprocessing, model definitions, training loops, evaluation metrics, and multiple experiment scripts for analysis.
Merit describes the code at this commit. Later pushes count once the owner rescans. How merit is scored
How much of the project is about each field, as scoring counts it. It ranks in the field it is most about.
Read from the project's code at its scored commit.
The owner keeps this project's limitations private.
