Question about the eye region

[zhoutianyang2002]

Dear authors,

Thanks for your excellent work!
I am trying to run LAM on the testset of VFHQ. I organized the motion-sequence tracking code based on the project's existing code.
Here is the final inference result, the eye region seems not good.

I check the tracking result, it seems no problem.

I do confused about the result. What should I do next?
Besides, may I ask if there is any ready-made inference code for the VFHQ dataset? I would be very grateful if you could share this part of the code.

Thank you very much!