Modeling Mutual Context of Object and Human Pose in Human-Object Interaction Activities