UDNet: Uncertainty-aware deep network for salient object detection

作者：

Highlights：

• We propose an uncertainty-aware SOD model with a high-level contrast feature extraction module and three decoding branches.

• We design a feature interaction module to enhance the feature interaction among internal contour uncertainty features, saliency features and external contour uncertainty features.

• Extensive experiments on four public benchmark datasets and the challenging SOC dataset demonstrate the superiority of the proposed model.

摘要

•We propose an uncertainty-aware SOD model with a high-level contrast feature extraction module and three decoding branches.•We design a feature interaction module to enhance the feature interaction among internal contour uncertainty features, saliency features and external contour uncertainty features.•Extensive experiments on four public benchmark datasets and the challenging SOC dataset demonstrate the superiority of the proposed model.

论文关键词：Salient object detection,Contour uncertainty,Feature interaction

论文评审过程：Received 28 March 2022, Revised 17 September 2022, Accepted 4 October 2022, Available online 5 October 2022, Version of Record 13 October 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.109099