For pose estimation. Should probably just take the keypoint of reference and subtract that from all other points. Maybe with an option to keep allocentric coordinates for the chosen keypoint so it's possible to convert back?
Could also keep the animal heading if two keypoints are given, 0 being either in-line or perpendicular to the line formed by the two keypoints.
If no reference is given, then compute the centroid and keep allocentric coordinates for that.
For pose estimation. Should probably just take the keypoint of reference and subtract that from all other points. Maybe with an option to keep allocentric coordinates for the chosen keypoint so it's possible to convert back?
Could also keep the animal heading if two keypoints are given, 0 being either in-line or perpendicular to the line formed by the two keypoints.
If no reference is given, then compute the centroid and keep allocentric coordinates for that.