Research map / Label-efficient and robust learning / Domain adaptation and generalization
Uncertainty-Aware Pseudo Label Refinery for Domain Adaptive Semantic Segmentation
Yuxi Wang, Junran Peng, ZhaoXiang 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 |
|---|---|---|---|
| add adversarial adaptation (ablation) | source-only model · GTA5->Cityscapes | mIoU: 36.6 -> 42.9 | Improved |
| add image translation (ablation) | AA baseline · GTA5->Cityscapes | mIoU: 42.9 -> 45.3 | Improved |
| add soft-balance resampling on source domain (ablation) | AA+IT · GTA5->Cityscapes | mIoU: 45.3 -> 46.1 | Improved |
| add target-guided uncertainty rectifying adversarial adaptation (TGAA) (ablation) | AA+IT+SR · GTA5->Cityscapes | mIoU: 46.1 -> 48.8 | Improved |
| add uncertainty-aware pseudo-label self-training (UPST) (ablation) | UncerDA (AA) · GTA5->Cityscapes | mIoU: 48.8 -> 52.6 | Improved |
| use class-balance sampling instead of no balance (ablation) | multi-level adaptation baseline · GTA5->Cityscapes | mIoU: 42.9 -> 45.7 | Improved |
| use soft-balance sampling lambda 0.9 instead of class-balance (ablation) | multi-level adaptation baseline · GTA5->Cityscapes | mIoU: 45.7 -> 46.1 | Improved |
| use focal loss instead of no balance (ablation) | multi-level adaptation baseline · GTA5->Cityscapes | mIoU: 42.9 -> 44.5 | Improved |
| use uncertainty-aware pseudo labels instead of CBST (ablation) | AdaptSegNet · GTA5->Cityscapes | mIoU: 47.1 -> 47.8 | Improved |
| use uncertainty-aware pseudo labels instead of CBST (ablation) | MRNet · GTA5->Cityscapes | mIoU: 49.7 -> 51.2 | Improved |
| use uncertainty-aware pseudo labels instead of CBST (ablation) | UncerDA (AA) · GTA5->Cityscapes | mIoU: 51.2 -> 52.6 | Improved |
| vary top-k uncertain classes from 5 to 9 (ablation) | UncerDA · GTA5/SYNTHIA->Cityscapes | mIoU | No clear effect |
Closest papers by meaning
- Rectifying Pseudo Label Learning via Uncertainty Estimation for Domain Adaptive Semantic Segmentation (IJCV 2021 · similarity 0.95)
- Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021 · similarity 0.92)
- LabOR: Labeling Only if Required for Domain Adaptive Semantic Segmentation (ICCV 2021 · similarity 0.90)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.88)
- Towards Robust Pseudo-Label Learning in Semantic Segmentation: An Encoding Perspective (NeurIPS 2025 · similarity 0.88)
- Connectivity-Driven Pseudo-Labeling Makes Stronger Cross-Domain Segmenters (NeurIPS 2024 · similarity 0.87)
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.
