What is the purpose of using a Softmax function at the end of Decoder path of U$-$Net model?

To transform the output of the Decoder path into a probability distribution over different classes for the Segmentation task

The Softmax function helps to reduce the dimensions of the output by removing the noise captured via the skip connections

It creates a probability distribution for fine$-$grained segmentations learned by the model

It helps in normalizing the output to have zero mean and unit variance