Training model with different kinds of labels in Keras/tensorflow

Hi,
Is it possible to train a model with different kind of masks (like two masks for each image) for detection or segmentation problem using Keras/tensorflow