The masks of CelebAMask-HQ were manually-annotated with the size of 512 x 512 and 19 classes including all facial components and accessories such as skin, nose, eyes, eyebrows, ears, mouth, lip, hair, hat, eyeglass, earring, necklace, neck, and cloth.

• Invert residual blocks design was used for Squeeze and Exapansion.
• DepthWise Convolution was used for the balance of parameter efficiency and accuracy.
• gelu was used in upscale pathway for smooth graident.
• Trained with a Polynomial Decay Schedule.

https://github.com/zllrunning/face-parsing.PyTorch