OCVOS: Object-Centric Representation for Video Object Segmentation | IEEE Conference Publication | IEEE Xplore