Is there any multiclass loss function for segmentation using UNET in keras?

Question

I am experimenting with multiclass segmentation in UNET. Is there any loss function built exactly for this?