SU2GE-Net: a saliency-based approach for non-specific class foreground segmentation

Xiaochun Lei; Xiang Cai; Linjun Lu; Zihang Cui; Zetao Jiang

doi:10.1038/s41598-023-40175-9

SU²GE-Net: a saliency-based approach for non-specific class foreground segmentation

Sci Rep. 2023 Aug 15;13(1):13263. doi: 10.1038/s41598-023-40175-9.

Authors

Xiaochun Lei^#^{1

2}, Xiang Cai^#¹, Linjun Lu¹, Zihang Cui¹, Zetao Jiang^{3

4}

Affiliations

¹ School of Computer Science and Information Security, Guilin University of Electronic Technology, GuiLin, 541010, Guangxi, China.
² Guangxi Key Laboratory of Image and Graphic Intelligent Processing, Guilin University of Electronic Technology, Guilin, 541004, Guangxi, China.
³ School of Computer Science and Information Security, Guilin University of Electronic Technology, GuiLin, 541010, Guangxi, China. zetaojiang@guet.edu.cn.
⁴ Guangxi Key Laboratory of Image and Graphic Intelligent Processing, Guilin University of Electronic Technology, Guilin, 541004, Guangxi, China. zetaojiang@guet.edu.cn.

^# Contributed equally.

Abstract

Salient object detection is vital for non-specific class subject segmentation in computer vision applications. However, accurately segmenting foreground subjects with complex backgrounds and intricate boundaries remains a challenge for existing methods. To address these limitations, our study proposes SU²GE-Net, which introduces several novel improvements. We replace the traditional CNN-based backbone with the transformer-based Swin-TransformerV2, known for its effectiveness in capturing long-range dependencies and rich contextual information. To tackle under and over-attention phenomena, we introduce Gated Channel Transformation (GCT). Furthermore, we adopted an edge-based loss (Edge Loss) for network training to capture spatial-wise structural details. Additionally, we propose Training-only Augmentation Loss (TTA Loss) to enhance spatial stability using augmented data. Our method is evaluated using six common datasets, achieving an impressive [Formula: see text] score of 0.883 on DUTS-TE. Compared with other models, SU²GE-Net demonstrates excellent performance in various segmentation scenarios.