Research map / Detection and segmentation / CNN features
Enhanced Boundary Learning for Glass-Like Object Segmentation
Hao He, Xiangtai Li, Guangliang Cheng, Jianping Shi, Yunhai Tong, Gaofeng Meng, Véronique Prinet, LuBin Weng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Refined Differential Module (RDM) (ablation) | Deeplabv3+ · Trans10k val | mIoU: 85.4 -> 89.2 | Improved |
| add cascade of RDM modules (ablation) | Deeplabv3+ + RDM · Trans10k val | mIoU: 89.2 -> 90.1 | Improved |
| add Point-based GCN Module (PGM) (ablation) | Deeplabv3+ + cascaded RDM · Trans10k val | mIoU: 90.1 -> 90.7 | Improved |
| add BCE edge loss supervision (ablation) | L_merge only · Trans10k val | mIoU: 88.3 -> 89.4 | Improved |
| add residual-part loss L_residual (ablation) | merge + BCE edge loss · Trans10k val | mIoU: 89.4 -> 90.4 | Improved |
| use Dice loss instead of BCE for edge loss (ablation) | EBLNet losses · Trans10k val | mIoU: 90.4 -> 90.7 | Improved |
| add low-level feature F_low in RDM (ablation) | RDM · Trans10k val | mIoU: 88.8 -> 89.2 | Improved |
| add high-level feature F_high in RDM (ablation) | RDM · Trans10k val | mIoU: 88.4 -> 89.2 | Improved |
| add subtraction and residual learning in RDM (ablation) | RDM · Trans10k val | mIoU: 86.4 -> 89.2 | Improved |
| append RDM+PGM modules to FCN | FCN · Trans10k val | mIoU: 83.2 -> 87.9 | Improved |
| append RDM+PGM modules to PSPNet | PSPNet · Trans10k val | mIoU: 83.5 -> 88.6 | Improved |
| append RDM+PGM modules to DANet | DANet · Trans10k val | mIoU: 84.1 -> 87.8 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DANet — DANet: Divergent Activation for Weakly Supervised Object Localization (ICCV 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Large-Field Contextual Feature Learning for Glass Detection (TPAMI 2022 · similarity 0.89)
- Segmenting Transparent Objects in the Wild (ECCV 2020 · similarity 0.88)
- Exploiting Semantic Relations for Glass Surface Detection (NeurIPS 2022 · similarity 0.88)
- Rich Context Aggregation With Reflection Prior for Glass Surface Detection (CVPR 2021 · similarity 0.88)
- Leveraging RGB-D Data with Cross-Modal Context Mining for Glass Surface Detection (AAAI 2025 · similarity 0.86)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
