In my understanding you use a well-trained expression regressor (R_exp) to conduct this constraint. So I'd like to know whether the regressor (R_exp) is trained differently according to the style, like for anime style you may train R_exp on anime images with anime face landmark detectors, but for 3d cartoon style you may train R_exp on 3d cartoon images with 3d cartoon face landmark detectors? Is that so? Wish your reply.
In my understanding you use a well-trained expression regressor (R_exp) to conduct this constraint. So I'd like to know whether the regressor (R_exp) is trained differently according to the style, like for anime style you may train R_exp on anime images with anime face landmark detectors, but for 3d cartoon style you may train R_exp on 3d cartoon images with 3d cartoon face landmark detectors? Is that so? Wish your reply.