Abstract: Existing leading Co-saliency Detection (CoD) framework aims to segment the co-salient objects by learning the consensus visual representation of the foreground objects. However, despite ...