Open Angel113110 opened 4 years ago
Hi,
Can you post some images of the unexpected results you get? You have to keep in mind that there will be errors, as the estimation is approximate.
For example, I attach two pictures for illustrating. The picture 7's value is larger than the picture 11. But, you can find that the person in picture 7 is looking forward and 11 is looking side. The up one is picture 7 and the other is 11.
What command line arguments did you use when extracting head pose from the images? When I try them, they look fine:
I'm using:
FaceLandmarkImg -multi_view 1 -wild -fdir <location of images>
Dear author I use FeatureExtraction -fdir -pose to obtain head pose results. I have tried the solution you offered, but, when a person look side it looks invalid. I offer an example as follows:
The pictures above seems wrong in -Pose_Rx value. Can you help to solve it?
Can you send the complete profile image without the detected landmarks? The bounding box looks quite wrong in that.
With regards to the three images with pitch variations (pose_Rx), two of them seem to be roughly correct while only one of them is obviously wrong (further OpenFace knows that one of them is wrong, as the confidence value assigned for that detection is really low)
@Angel113110 from the openface output values (in the csv) how do you detect if the person is looking up, down, right or left? Thank you
@Angel113110 @TadasBaltrusaitis so, i obtained this values (namely gaze_angle, pose_T and so on) but how can i use them to understand if looking right, left, down or up? What are the intervals to determine if person looks right, left, up or down?
Dear author Recently, I tried to use openface to do head pose estimation. I met an problem which confuse me. That is some pictures result value in which the person's head is more partial than others is smaller than other pictures in which the head is straight. That's weird. Can you help me? Thanks!