Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain-generalized segmentation
Generalizes segmentation to unseen domains using style augmentation, memory and robust foundation-model tuning.
56 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Domain adaptation and generalization, keyword cluster “segmentation · semantic · domain adaptive”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
15 papers in 2023–24 and 22 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 4.4% to 6.9%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Evaluation of Semantic Dependency Labeling Across Domains (AAAI 2016 · 1 citation)
- 2018 · Fully Convolutional Adaptation Networks for Semantic Segmentation (CVPR 2018 · 372 citations)
- 2019 · DLOW: Domain Flow for Adaptation and Generalization (CVPR 2019 · 375 citations)
- 2020 · Generalizing Hand Segmentation in Egocentric Videos With Uncertainty-Guided Model Adaptation (CVPR 2020 · 51 citations)
- 2021 · Dynamic to Static Lidar Scan Reconstruction Using Adversarially Trained Auto Encoder (AAAI 2021 · 5 citations)
- 2022 · WildNet: Learning Domain Generalized Semantic Segmentation From the Wild (CVPR 2022 · 110 citations)
- 2023 · Single Domain Generalization for LiDAR Semantic Segmentation (CVPR 2023 · 43 citations)
- 2024 · Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · 33 citations)
- 2025 · FisherTune: Fisher-Guided Robust Tuning of Vision Foundation Models for Domain Generalized Segmentation (CVPR 2025 · 16 citations)
- 2026 · Earth-Adapter: Bridge the Geospatial Domain Gaps with a Frequency-Guided Mixture of Adapters (AAAI 2026 · 3 citations)
Earliest papers here
- Evaluation of Semantic Dependency Labeling Across Domains (AAAI 2016 · 1 citation)
- Fully Convolutional Adaptation Networks for Semantic Segmentation (CVPR 2018 · 372 citations)
- Conditional Generative Adversarial Network for Structured Domain Adaptation (CVPR 2018 · 347 citations)
What has been tried
37 papers here report 436 measured changes (a component added, removed or swapped, compared with a baseline): helped 83%, no clear effect 4%, hurt 12%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add the feature perturbation module | CD-FSS 1-shot average · mIoU | 57.20 -> 58.45 | Helped | Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · 33 citations) |
| Add covariance matching loss (CML) | Cityscapes, BDD-100K, and Mapillary · mIoU (%) | 28.95 -> 38.08; 25.14 -> 36.65; 28.18 -> 40.62 | Helped | Style Blind Domain Generalized Semantic Segmentation via Covariance Alignment and Semantic Consistence Contrastive Learning (CVPR 2024 · 33 citations) |
| add a Bridge Domain | SemanticKITTI→SemanticSTF · mIoU | 31.6 -> 37.2 | Helped | UniMix: Towards Domain Adaptive and Generalizable LiDAR Semantic Segmentation in Adverse Weather (CVPR 2024 · 32 citations) |
| Add style encoder S to content encoder C | ACDC-fog · mIoU | 73.4 -> 75.2 | Helped | Learning Generalized Segmentation for Foggy-Scenes by Bi-directional Wavelet Guidance (AAAI 2024 · 31 citations) |
| Add low-frequency LL adaptation | CityScapes → BDD-100K · mIoU (%) | 62.43 -> 63.85 | Helped | Learning Frequency-Adapted Vision Foundation Model for Domain Generalized Semantic Segmentation (NeurIPS 2024 · 25 citations) |
| select 5% of parameters using DR-FIM rather than random selection | Cityscapes → BDD100K · mIoU | 61.1 -> 65.8 | Helped | FisherTune: Fisher-Guided Robust Tuning of Vision Foundation Models for Domain Generalized Segmentation (CVPR 2025 · 16 citations) |
| Add HRFP+ and NP+ rather than use HRFP+ alone | BDD-100k, Cityscapes, Mapillary, and Synthia val (G→target) · mIoU | 39.28 -> 39.55; 41.39 -> 42.40; 42.70 -> 44.93; 29.70 -> 30.22 | Helped | MRFP: Learning Generalizable Semantic Segmentation from Sim-2-Real with Multi-Resolution Feature Perturbation (CVPR 2024 · 14 citations) |
| Add brightness sampling to the HRDA baseline | ACDC-All · mIoU | 68.0 -> 68.8; +0.8 | Helped | Parsing All Adverse Scenes: Severity-Aware Semantic Segmentation with Mask-Enhanced Cross-Domain Consistency (AAAI 2024 · 13 citations) |
| Add SQB to the baseline | GTAV → Cityscapes, BDD-100K, Mapillary, SYNTHIA (Avg.) · mIoU (%) | 46.41 -> 47.43 (+1.02) | Helped | Exploring Semantic Consistency and Style Diversity for Domain Generalized Semantic Segmentation (AAAI 2025 · 10 citations) |
| Add IBA to the REIN baseline | AN, DZ, ND · mIoU | +1.9%, +3.5%, +2.4% | Helped | NightAdapter: Learning a Frequency Adapter for Generalizable Night-time Scene Segmentation (CVPR 2025 · 9 citations) |
| add ETI reconstructed images and hybrid pseudo labels | DSEC-Semantic · mIoU | 44.87 -> 51.08 | Helped | HPL-ESS: Hybrid Pseudo-Labeling for Unsupervised Event-based Semantic Segmentation (CVPR 2024 · 8 citations) |
| Add APM-S to the HSNet baseline | CD-FSS average, 1-shot · mIoU | 47.68 -> 57.77 | Helped | Lightweight Frequency Masker for Cross-Domain Few-Shot Semantic Segmentation (NeurIPS 2024 · 7 citations) |
What reviewers will expect
Datasets and baselines named in the 15 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Cityscapes · 11 of 15 papers
- SYNTHIA · 6 of 15 papers
- BDD-100K · 6 of 15 papers
- Mapillary · 6 of 15 papers
- ACDC · 5 of 15 papers
- GTAV · 5 of 15 papers
- FSS-1000 · 2 of 15 papers
- DeepGlobe · 2 of 15 papers
Baselines
- WildNet · 5 of 15 papers
- ISW · 4 of 15 papers
- SHADE · 4 of 15 papers
- HRDA · 4 of 15 papers
- Iternorm · 3 of 15 papers
- SPC · 3 of 15 papers
- HGFormer · 3 of 15 papers
- DAFormer · 3 of 15 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 37 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 2 papers
- University of Trento · 2 papers
- Xidian University · 2 papers
Use it
Download BibTeX of 35 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Adaptation for semantic segmentation (81 papers)
- Cross-modality medical segmentation adaptation (20 papers)
Elsewhere in the topic
- Conditional distribution alignment (162 papers)
- Cross-domain few-shot with transformers (116 papers)
- Feature alignment for unsupervised adaptation (105 papers)
- Optimization-based domain generalization (101 papers)
- Adversarial domain adaptation (95 papers)
- Geometric alignment and synthetic-to-real adaptation (93 papers)
- Source-free domain adaptation (92 papers)
- Statistical alignment and hypothesis testing (80 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).