CSRNet: Cascaded Selective Resolution Network for real-time semantic segmentation. (January 2023)

• Record Type: Journal Article
• Authors: Xiong, Jingjing; Po, Lai-Man; Yu, Wing-Yin; Zhou, Chang; Xian, Pengfei; Ou, Weifeng
• Is Part Of: Expert systems with applications. Volume 211(2023)
• Publication Date: 2023-01
• Subjects: Semantic segmentation -- Attention mechanism -- Real-time inference -- Deep neural networks; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.118537
• Languages: English
• ISSNs: 0957-4174

Abstract

Abstract: Real-time semantic segmentation has received considerable attention due to growing demands in many practical applications, such as autonomous vehicles, robotics, etc. Existing real-time segmentation approaches often utilize feature fusion to improve segmentation accuracy. However, they fail to fully consider the feature information at different resolutions and the receptive fields of the networks are relatively limited, thereby compromising the performance. To tackle this problem, we propose a light Cascaded Selective Resolution Network (CSRNet) to improve the performance of real-time segmentation through multiple context information embedding and enhanced feature aggregation. The proposed network builds a three-stage segmentation system, which integrates feature information from low resolution to high resolution and achieves feature refinement progressively. CSRNet contains two critical modules: the Shorted Pyramid Fusion Module (SPFM) and the Selective Resolution Module (SRM). The SPFM is a computationally efficient module to incorporate the global context information and significantly enlarge the receptive field at each stage. The SRM is designed to fuse multi-resolution feature maps with various receptive fields, which assigns soft channel attentions across the feature maps and helps to remedy the problem caused by multi-scale objects. Comprehensive experiments on well-known road scene datasets demonstrate that the proposed CSRNet outperforms the main-stream efficient semantic segmentation approaches by accuracy and can be performed at a fast inference speed. Highlights: Multiple-stage segmentation network refines the feature maps progressively. Multiple context information embedding enlarges the receptive field at each stage. Selective Resolution Module aggregates multi-resolution feature maps. The proposed network alleviates the problem caused by multi-scale objects. The proposed system shows improved accuracy for real-time segmentation.