I am using the attention tracker package. It maps points on the face of a 2D image to a 3D model of the head to generate a homogeneous transform matrix pose to describe the camera to head transformation. Using pose, and the points (0,0,0), (50,0,0), (0,50,0), (0,0,50) the debugging feature maps the points back onto the image to get an idea of what the coordinate axis looks like. The coordinate axis on the head looks good (satisfies the ROS REP 103 axis orientation standards), this is how I want the coordinates to look. However, when I took a look at the published tf transform in RVIZ the head coordinate axis is upside down and the movements inverted (compared to those displayed by the debug image).