Research map / Label-efficient and robust learning / Domain adaptation and generalization
Cluster, Split, Fuse, and Update: Meta-Learning for Open Compound Domain Adaptive Semantic Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 cluster and split modules to AdaptSegNet | GTA5→BDD100K segmentation · BDD100K target domain | mIoU: +3.1% | Helped |
| add cluster and split modules to OCDA | GTA5→BDD100K segmentation · BDD100K target domain | mIoU: +0.9% | Helped |
| add adaptive hypernetwork fusion and MAML training to the split model (ablation) | MOCDA · GTA5→BDD100K target domain | mIoU: 25.4% to 27.7% | Helped |
| enable online update at inference (ablation) | MOCDA · GTA5→BDD100K open domains average | mIoU: 28.1% to 28.8%; +0.7 | Helped |
| enable online update at inference (ablation) | MOCDA · GTA5→BDD100K overcast | mIoU: 31.4 -> 31.4 | No clear effect |
| enable online update at inference (ablation) | MOCDA · SYNTHIA-SF→BDD100K open domains average | mIoU: 30.1% to 31.2%; +1.1 | Helped |
| replace adaptive hypernetwork fusion with average fusion at testing only (ablation) | MOCDA · GTA5→BDD100K target domain | mIoU: 27.1 -> 23.1 | Hurt |
| replace adaptive hypernetwork fusion with average fusion during training and testing (ablation) | MOCDA · GTA5→BDD100K target domain | mIoU: 27.1 -> 26.1 | Hurt |
| replace adaptive hypernetwork fusion with style-code-distance weighting (ablation) | MOCDA · GTA5→BDD100K target domain | mIoU: 27.1 -> 26.6 | Hurt |
| add unsupervised entropy loss (ablation) | MOCDA with hypernetwork fusion · GTA5→BDD100K target domain | mIoU: 27.1% to 27.3% | Helped |
| add MAML training to the fuse module (ablation) | MOCDA with entropy loss · GTA5→BDD100K target domain | mIoU: 27.3 -> 27.7 | Helped |
| add cluster and split modules | AdaptSegNet · SYNTHIA-SF→BDD100K | mIoU: +2.4% target domain; +2.3% open domain | Helped |
Closest papers by meaning
- Discover, Hallucinate, and Adapt: Open Compound Domain Adaptation for Semantic Segmentation (NeurIPS 2020 · similarity 0.92)
- Learning to Adapt via Latent Domains for Adaptive Semantic Segmentation (NeurIPS 2021 · similarity 0.91)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · similarity 0.88)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.87)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Towards Unsupervised Domain Bridging via Image Degradation in Semantic Segmentation (NeurIPS 2025 · similarity 0.86)
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.