Engaging Part-Whole Hierarchies and Contrast Cues for Salient Object Detection

Real-world scenes always exhibit objects with clutter backgrounds, posing great challenges for deep salient object detection models. In this paper, we propose by engaging two saliency cues, i.e. , the part-whole hierarchies and contrast resulting in a PWHCNet. Specifically, branches, which consists of Dynamic Grouping Capsules (DGC) branch DenseHRNet branch, are put place to learn respectively. Moreover, help highlight whole complex scenes, Background Suppression (BS) module is proposed guide shallow features aid relational cues captured DGC. Subsequently, these integrated via Self-Channel Mutual-Spatial (SCMS) attention mechanism. Experimental results on five benchmarks demonstrate that PWHCNet achieves state-of-the-art performance while obtaining fine details.