Incomplete segmentation results of Celebm model inference faceparsing

[winner-4]

Is there any method to solve the problem of incomplete segmentation when inferring faceparsing results using the Celebm model provided by the author?

[winner-4]

[XavierJiezou]

It's my solution: https://github.com/FacePerceiver/facer/issues/36

[ChaoLi977]

It's my solution: #36

I think the solution is not the answer of the incomplete segmentation issue.