A Deep Learning Framework for Boundary-Aware Semantic Segmentation

2025-03-28

AI Summary

To address ambiguous object boundaries and poor small-object segmentation in semantic segmentation, this paper introduces the Boundary-Enhanced Feature Bridging Module (BEFBM) into the Mask2Former framework—the first such integration. BEFBM explicitly models boundary geometry to construct boundary-aware feature maps and enables cross-scale fusion of semantic and boundary features. It employs joint semantic-boundary supervision to strengthen fine-grained boundary preservation and segmentation consistency. Experiments on Cityscapes demonstrate consistent improvements: +2.3% mIoU, +3.1% mDice, and +4.5% mRecall, significantly enhancing boundary fidelity and small-object segmentation accuracy in complex scenes. This work establishes a scalable, boundary-aware modeling paradigm for transformer-based segmentation architectures.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Feature Construction/Reformulation
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

As a fundamental task in computer vision, semantic segmentation is widely applied in fields such as autonomous driving, remote sensing image analysis, and medical image processing. In recent years, Transformer-based segmentation methods have demonstrated strong performance in global feature modeling. However, they still struggle with blurred target boundaries and insufficient recognition of small targets. To address these issues, this study proposes a Mask2Former-based semantic segmentation algorithm incorporating a boundary enhancement feature bridging module (BEFBM). The goal is to improve target boundary accuracy and segmentation consistency. Built upon the Mask2Former framework, this method constructs a boundary-aware feature map and introduces a feature bridging mechanism. This enables effective cross-scale feature fusion, enhancing the model's ability to focus on target boundaries. Experiments on the Cityscapes dataset demonstrate that, compared to mainstream segmentation methods, the proposed approach achieves significant improvements in metrics such as mIOU, mDICE, and mRecall. It also exhibits superior boundary retention in complex scenes. Visual analysis further confirms the model's advantages in fine-grained regions. Future research will focus on optimizing computational efficiency and exploring its potential in other high-precision segmentation tasks.

Problem

• Improves target boundary accuracy in semantic segmentation
• Enhances recognition of small targets in segmentation tasks
• Addresses blurred boundaries in Transformer-based segmentation methods

Innovation

• Boundary enhancement feature bridging module (BEFBM)
• Mask2Former-based semantic segmentation algorithm
• Cross-scale feature fusion for boundary focus