Multi-scale iterative refinement network for RGB-D salient object detection

The extensive research leveraging RGB-D information has been exploited in salient object detection. However, visual cues appear various scales and resolutions of RGB images due to semantic gaps at different feature levels. Meanwhile, similar patterns are available cross-modal depth as well multi-scale versions. Cross-modal fusion refinement still an open problem detection task. In this paper, we begin by introducing top-down bottom-up iterative architecture leverage features, then devise attention based module (ABF) address on correlation. We conduct experiments seven public datasets. experimental results show the effectiveness our devised method