Why does my keras image classification model produce overconfident results?

Question

I have a problem with to confident outcomes of a Keras model. So my dataset consists of 6000 Images divided into 8 classes. Validation = 20% of the dataset. The architecture loo