Calibrating probabilities of neural network

Question

I would like to calibrate the probabilites of my CNN classification model. I have a custom cross-validation and in each cross-validation run I\'m dowing the following:

<