Research map / Label-efficient and robust learning / Domain adaptation and generalization
Union-Set Multi-source Model Adaptation for Semantic Segmentation
Zongyao Li, Ren Togo, Takahiro Ogawa, Miki Haseyama
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 cross-model consistency to self-training (ablation) | ST · Cityscapes (non-overlapping, S+G) | mIoU: 42.3 -> 43.2 | Improved |
| add adversarial learning on top of cross-model consistency (ablation) | ST+CMC · Cityscapes (non-overlapping, S+G) | mIoU: 43.2 -> 44.0 | Improved |
| add maximum squares loss (ablation) | ST+CMC+ADV · Cityscapes (non-overlapping, S+G) | mIoU: 44.0 -> 45.8 | Improved |
| add model integration by distillation (Stage II) (ablation) | Stage I of PM · Cityscapes (non-overlapping, S+G) | mIoU: 45.8 -> 46.6 | Improved |
| add cross-model consistency to self-training (ablation) | ST · Cityscapes (partly-overlapping, S+G) | mIoU: 44.2 -> 46.0 | Improved |
| add maximum squares loss (ablation) | ST+CMC+ADV · Cityscapes (partly-overlapping, S+G) | mIoU: 46.6 -> 47.4 | Improved |
| add model integration (ablation) | Stage I of PM · Cityscapes (partly-overlapping, S+G) | mIoU: 47.4 -> 48.3 | Improved |
| add model-invariant feature learning to common classes of S+T (ablation) | ST · Cityscapes (partly-overlapping, S+T, common classes) | mIoU: 49.6 -> 49.8 | No clear effect |
| add maximum squares loss (ablation) | ST+CMC+ADV · Cityscapes (fully-overlapping, S+G) | mIoU: 50.9 -> 51.0 | No clear effect |
| add model integration (ablation) | Stage I of PM · Cityscapes (fully-overlapping, S+G) | mIoU: 51.0 -> 51.7 | Improved |
| use proposed method instead of weighted-ensemble MSMA | MSMA · Cityscapes (non-overlapping, S+G) | mIoU: 30.2 -> 46.6 | Improved |
| use proposed method (Stage I only) instead of SSMA/UDA | Stage I of PM · Cityscapes (partly-overlapping, S+G) | mIoU: 47.4 -> 47.9 | Hurt |
Closest papers by meaning
- Multi-Source Domain Adaptation With Collaborative Learning for Semantic Segmentation (CVPR 2021 · similarity 0.90)
- Unsupervised Model Adaptation for Continual Semantic Segmentation (AAAI 2021 · similarity 0.89)
- Generalize Then Adapt: Source-Free Domain Adaptive Semantic Segmentation (ICCV 2021 · similarity 0.89)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.88)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 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.
