In this paper, we propose a separable structure modeling approach for semi-supervised video object segmentation. Unlike most existing methods which preclude the semantically structural information of ...