Abstract: How to effectively model both spatial information and temporal dynamics is crucial to Video Salient Object Detection (VSOD). Recently, there are some works using self-attention mechanism to ...