CoADNet: Collaborative Aggregation-and-Distribution Networks for Co-Salient Object Detection

Co-Salient Object Detection (CoSOD) aims at discovering salient objects that\nrepeatedly appear in a given query group containing two or more relevant\nimages. One challenging issue is how to effectively capture co-saliency cues by\nmodeling and exploiting inter-image relationships. In this paper, we present an\nend-to-end collaborative aggregation-and-distribution network (CoADNet) to\ncapture both salient and repetitive visual patterns from multiple images.\nFirst, we integrate saliency priors into the backbone features to suppress the\nredundant background information through an online intra-saliency guidance\nstructure. After that, we design a two-stage aggregate-and-distribute\narchitecture to explore group-wise semantic interactions and produce the\nco-saliency features. In the first stage, we propose a group-attentional\nsemantic aggregation module that models inter-image relationships to generate\nthe group-wise semantic representations. In the second stage, we propose a\ngated group distribution module that adaptively distributes the learned group\nsemantics to different individuals in a dynamic gating mechanism. Finally, we\ndevelop a group consistency preserving decoder tailored for the CoSOD task,\nwhich maintains group constraints during feature decoding to predict more\nconsistent full-resolution co-saliency maps. The proposed CoADNet is evaluated\non four prevailing CoSOD benchmark datasets, which demonstrates the remarkable\nperformance improvement over ten state-of-the-art competitors.\n