Keras custom loss function with multiple output model

Question

In a segmentation task I wanted to have my model to have two outputs because I implemented weight maps as suggested in the original U-net paper https://arxiv.org/pdf/1505.04