cv::aruco::estimatePose returns the mapping between marker space and camera space via the two variables rvec and tvec - but the documentation doesn't state how exactly. In particular I wonder whether it describes camera location/rotation in the marker space or whether it describes marker location/rotation in camera space?
Or stated differently, is it
point_in_marker_space = (Rodrigues(rvec) * point_in_camera_space) + tvec
or
point_in_camera_space = (Rodrigues(rvec) * point_in_marker_space ) + tvec
or something else?