Visual cues for view-invariant action recognition