Change Loss Function in vgg16

Hey guys, I’m using vgg16 net to perform semantic segmentation on 2D images.

Is there a way I can fine-tine this network (vgg16) and change the loss function? An example would be great.

Thanks, David