A student of @hjchang reported the following issue:
When I cover half of my face with my hand, rt_gene gives me an incorrect head pose estimation. I guess the cause of the problem is that the program recorded the rotation matrix and translation matrix of the last time, because when I occlude half of the face, the face recognition module will still detect the face normally, but will give a face landmark with large error (See the attached picture for details). At this time, if we use this face landmark to calculate the rotation matrix and translation matrix, we will get the wrong answer, and it will influence the next round calculation.
A student of @hjchang reported the following issue:
He expects the issue in the following code: https://github.com/Tobias-Fischer/rt_gene/blob/master/rt_gene/src/rt_gene/extract_landmarks_method.py#L335-L347
/cc @Twarz @ngageorange