face segmentation is the task of densely labeling pixels on the face
according to their semantics. While current methods place an emphasis on
developing sophisticated architectures, use conditional random fields for
smoothness, or rather employ adversarial training, we follow an altern