philgras/video-head-tracker

About the FLAMEHead Layer

Opened this issue · 0 comments

SuperStacie commented

Hi,

Thanks for this great and useful module!
I noticed there are some differences in your FLAMEHead layer compared with DECA, EMOCA, etc. I wonder the reason:

  1. why the jaw pose at neutral state is set to be as a very wide open mouth
  2. why apply the zero_center
  3. what will happen if the neutral poses are all set to zeros and not apply rotation limits
  4. If there is specific reason you using calibrate_camera and solve_pnp in opencv instead of those in Pytorch

Thanks for taking time!