BiSeNet-oriented context attention model for image semantic segmentation

Lin Teng¹ and Yulong Qiao¹

College of Information and Communication Engineering, Harbin Engineering University
Harbin 150001, China
tenglinheu@163.com, qiaoyulong@hrbeu.edu.cn

Abstract

When the traditional semantic segmentation model is adopted, the different feature importance of feature maps is ignored in the feature extraction stage, which results in the detail loss, and affects the segmentation effect. In this paper, we propose a BiSeNet-oriented context attention model for image semantic segmenta-tion. In the BiSeNet, the spatial path is utilized to extract more low-level features to solve the problem of information loss in deep network layers. Context attention mechanism is used to mine high-level implied semantic features of images. Mean-while, the focus loss is used as the loss function to improve the final segmentation effect by reducing the internal weighting. Finally, we conduct experiments on open data sets, and the results show that pixel accuracy, average pixel accuracy, and aver-age Intersection-over-Union are greatly improved compared with other state-of-the-art semantic segmentation models. It effectively improves the accuracy of feature extraction, reduces the loss of feature details, and improves the final segmentation effect.

Key words

image semantic segmentation, BiSeNet, context attention, focus loss

Digital Object Identifier (DOI)

https://doi.org/10.2298/CSIS220321040T

Publication information

Volume 19, Issue 3 (September 2022)
Year of Publication: 2022
ISSN: 2406-1018 (Online)
Publisher: ComSIS Consortium

Full text

Download Available in PDF
Portable Document Format

How to cite

Teng, L., Qiao, Y.: BiSeNet-oriented context attention model for image semantic segmentation. Computer Science and Information Systems, Vol. 19, No. 3, 1409-1426. (2022), https://doi.org/10.2298/CSIS220321040T