Text this: Object, scene and ego-centric action classification for robotic vision /