Unified Information Fusion Network for Multi-Modal RGB-D and RGB-T Salient Object Detection

The use of complementary information, namely depth or thermal has shown its benefits to salient object detection (SOD) during recent years. However, the RGB-D RGB-T SOD problems are currently only solved independently, and most them directly extract fuse raw features from backbones. Such methods can be easily restricted by low-quality modality data redundant cross-modal features. In this work, a unified end-to-end framework is designed simultaneously analyze tasks. Specifically, effectively tackle multi-modal features, we propose novel multi-stage multi-scale fusion network (MMNet), which consists module (CMFM) bi-directional decoder (BMD). Similar visual color stage doctrine in human system (HVS), proposed CMFM aims explore important feature representations response stage, integrate into adversarial combination stage. Moreover, BMD learns multi-level fused capture both local global information objects, further boost performance. cross-modality analysis based on two-stage used for diverse Comprehensive experiments ( $\sim 92\text{K}$ image-pairs) demonstrate that method consistently outperforms other 21 state-of-the-art nine benchmark datasets. This validates our work well tasks with good generalization robustness, provides benchmark.