Research map / Detection and segmentation / CNN features
Distribution-Aware Margin Calibration for Semantic Segmentation in Images
Litao Yu, Zhibin Li, Min Gang Xu, Yongsheng Gao, Jiebo Luo, Jian Zhang
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 |
|---|---|---|---|
| replace cross-entropy with margin calibration | COPLE-Net · Robotic Instrument test | mIoU: 66.2 -> 72.5 | Improved |
| replace focal loss with margin calibration | COPLE-Net · Robotic Instrument test | mIoU: 69.5 -> 72.5 | Improved |
| use Dice loss as sole objective | COPLE-Net · Robotic Instrument test | mIoU: 66.2 -> 27.3 | Hurt |
| use Tversky loss as sole objective | COPLE-Net · Robotic Instrument test | mIoU: 66.2 -> 43.8 | Hurt |
| replace cross-entropy with margin calibration | DeepLab v3+ · COCO-Stuff 10K val | mIoU: 34.1 -> 35.5 | Improved |
| replace Lovasz-softmax with margin calibration | DeepLab v3+ · PASCAL VOC2012 val | mIoU: 78.5 -> 78.6 | No clear effect |
| fine-tune with margin calibration instead of cross-entropy | DeepLab v3+ · SceneParse150 val (flip+multi-scale) | mIoU: 39.1 -> 40.2 | Improved |
| fine-tune with margin calibration instead of cross-entropy | DeepLab v3+ · Cityscapes val (flip+multi-scale) | mIoU: 79.4 -> 81.1 | Improved |
| fine-tune with margin calibration instead of cross-entropy | DeepLab v3+ · BDD100K val (flip+multi-scale) | mIoU: 65.7 -> 66.1 | Improved |
| fine-tune with margin calibration instead of cross-entropy | DeepLab v3+ · Mapillary Vistas val (flip+multi-scale) | mIoU: 49.8 -> 51.1 | Improved |
| add flipping and multi-scale inference (ablation) | DeepLab v3+ margin calibration · SceneParse150 val | mIoU: 38.5 -> 40.2 | Improved |
| replace cross-entropy with margin calibration (no coarse pretrain) | DeepLab v3+ SEResNeXt-50 · Cityscapes test | IoU class: 80.5 -> 81.5 | Improved |
Closest papers by meaning
- The Lovász-Softmax Loss: A Tractable Surrogate for the Optimization of the Intersection-Over-Union Measure in Neural Networks (CVPR 2018 · similarity 0.90)
- RankSEG-RMA: An Efficient Segmentation Algorithm via Reciprocal Moment Approximation (NeurIPS 2025 · similarity 0.85)
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.84)
- MaSS13K: A Matting-level Semantic Segmentation Benchmark (CVPR 2025 · similarity 0.83)
- Open-Vocabulary Segmentation with Semantic-Assisted Calibration (CVPR 2024 · similarity 0.83)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.83)
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.
