Data valuation for CheXpert dataset

[ffeiland]

Hi there,

I would like to apply the data valuation methods (KNN-Shap, DShap, TMC-Shap, LOO-values) on the Chexpert dataset (https://stanfordmlgroup.github.io/competitions/chexpert/) in order to detect noisy labels. Therefore, the ResNet152 model should be fine as a baseline.
As I am quite new to machine learning I would like to ask for the best way to do this.

• How do the images (320x320 pixels) need to be pre-processed in order to run the calculations?
• How should the training and test data be split up?

Thanks in advance,

Regards,
Fabian