Is it possible to provide pupil-invisible ROS topics ( publish gaze points and scene images, even eye images) for human-robot interaction control.