search

610265158 / Peppa_Pig_Face_Landmark

A simple face detect and alignment method, which is easy and stable.
Apache License 2.0
530 stars 116 forks source link

The meaning of the “states” derived from this keypoint model #7

Closed hanans426 closed 4 years ago

hanans426 commented 5 years ago

hi，I don’t understand the meaning of the “states” derived from this keypoint model, and why is this variable not used？

610265158 commented 5 years ago

hi，I don’t understand the meaning of the “states” derived from this keypoint model, and why is this variable not used？

hi, it is 4 cls params,(left eye, right eye, mouth open, mouth open widely), as we all known that multitask learning can improve the face alignment algorithm.
However, these params is not that useful, for a face keypoint model.

hanans426 commented 5 years ago

hi,On the tf1 branch, why can't the "headpose" parameter obtained through keypoint model be directly used for the headpose prediction?

610265158 commented 5 years ago

hi,On the tf1 branch, why can't the "headpose" parameter obtained through keypoint model be directly used for the headpose prediction?

hi, You still could get the pose params from the model.

hanans426 commented 5 years ago

hi，The “headpose” parameters obtained in the model, what is the order of “yaw,roll and pitch”? And why recalculate the head rotation Angle instead of just using these parameters obtained in the “keypoint” model?

修改翻译结果

610265158 commented 5 years ago

hi，The “headpose” parameters obtained in the model, what is the order of “yaw,roll and pitch”? And why recalculate the head rotation Angle instead of just using these parameters obtained in the “keypoint” model?

修改翻译结果 hi， sorry， i forget the order of the three params, but it should be the same with the one that recalculate. I think you can give it a test. There is no specially reason that recalculate the headpose, and there should be littel difference.

hanans426 commented 5 years ago

ok thank you so much