RRNet: Relational Reasoning Network With Parallel Multiscale Attention for Salient Object Detection in Optical Remote Sensing Images

Salient object detection (SOD) for optical remote sensing images (RSIs) aims at locating and extracting visually distinctive objects/regions from the optical RSIs. Since some saliency models were proposed to solve the intrinsic problem of optical RSIs (such as complex background and scale-variant objects), the accuracy and completeness are still unsatisfactory. To this end, we propose a relational reasoning network (RRNet) with parallel multiscale attention (PMA) for SOD in optical RSIs in this article. The relational reasoning module that integrates the spatial and the channel dimensions is designed to infer the semantic relationship by utilizing high-level encoder features, thereby promoting the generation of more complete detection results. The PMA module is proposed to effectively restore the detailed information and address the scale variation of salient objects by using the low-level features refined by multiscale attention. Extensive experiments on two datasets demonstrate that our proposed RRNet outperforms the existing state-of-the-art SOD competitors both qualitatively and quantitatively (https://rmcong.github.io/proj_RRNet.html).

Automated summary

This article proposes a relational reasoning network (RRNet) with parallel multiscale attention (PMA) for salient object detection (SOD) in optical remote sensing images (RSIs). By integrating the spatial and channel dimensions and utilizing high-level encoder features, the RRNet infers semantic relationships and generates more complete detection results. The PMA module effectively restores detailed information and addresses scale variation of salient objects.