Research map / Label-efficient and robust learning / Domain adaptation and generalization
Style Mixing and Patchwise Prototypical Matching for One-Shot Unsupervised Domain Adaptive Semantic Segmentation
Read the paperCheck 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 |
|---|---|---|---|
| Use the proposed style-mixing segmentor and patchwise prototypical matching | One-shot domain adaptation for semantic segmentation · GTA5 → Cityscapes-val | mIoU: 36.6 -> 42.8 | Helped |
| Use the proposed style-mixing segmentor and patchwise prototypical matching | One-shot domain adaptation for semantic segmentation · SYNTHIA → Cityscapes-val | mIoU: 38.6 -> 47.3 | Helped |
| Adapt using the proposed method | Cityscapes-trained segmentor · Cityscapes → Dark Zurich-val | mIoU: 14.6 -> 17.5 | Helped |
| Add the PPM confidence map to the loss (ablation) | PPM loss with standard source segmentation model · GTA5 → Cityscapes | mIoU: 38.38 -> 41.04 | Helped |
| Add the PPM confidence map to the loss (ablation) | PPM loss with standard source segmentation model · SYNTHIA → Cityscapes | mIoU: 34.26 -> 43.92 | Helped |
| Add entropy weighting E (ablation) | PPM loss with standard source segmentation model · GTA5 → Cityscapes | mIoU: 40.95 -> 41.04 | No clear effect |
| Add entropy weighting E (ablation) | PPM loss with standard source segmentation model · SYNTHIA → Cityscapes | mIoU: 44.14 -> 43.92 | Hurt |
| Add style mixing at image level and feature level f3 (ablation) | DeepLabV2-Res101 with PPM · GTA5 → Cityscapes | mIoU: 41.04 -> 42.77 | Helped |
| Add style mixing at image level and feature level f3 (ablation) | DeepLabV2-Res101 with PPM · SYNTHIA → Cityscapes | mIoU: 43.92 -> 47.33 | Helped |
| Use the modified style-mixing operation rather than original AdaIN (ablation) | Style-mixing segmentor · GTA5 → Cityscapes | mIoU: 41.98 -> 42.77 | Helped |
| Use the modified style-mixing operation rather than original AdaIN (ablation) | Style-mixing segmentor · SYNTHIA → Cityscapes | mIoU: 46.82 -> 47.33 | Helped |
| Set PPM patch size to 32 (ablation) | Patchwise prototypical matching · GTA5 → Cityscapes | mIoU: 42.26 -> 42.77 | Helped |
Closest papers by meaning
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.91)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Unsupervised Model Adaptation for Continual Semantic Segmentation (AAAI 2021 · similarity 0.90)
- Consistency Regularization with High-dimensional Non-adversarial Source-guided Perturbation for Unsupervised Domain Adaptation in Segmentation (AAAI 2021 · similarity 0.90)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.89)
- Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.89)
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.