Research map / Label-efficient and robust learning / Domain adaptation and generalization
Consistency Regularization with High-dimensional Non-adversarial Source-guided Perturbation for Unsupervised Domain Adaptation in 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 bidirectional style translation with source-guided unsupervised learning | BiSIDA · GTA5-to-CityScapes | mIoU: 42.5 -> 43.2 | Helped |
| use bidirectional style translation with source-guided unsupervised learning | BiSIDA · SYNTHIA-to-CityScapes | mIoU: 40.5 -> 42.1 | Helped |
| add target-guided source-to-target style translation (ablation) | BiSIDA · GTA5-to-CityScapes | mIoU: 29.3 -> 34.7 | Helped |
| add target-guided source-to-target style translation (ablation) | BiSIDA · SYNTHIA-to-CityScapes | mIoU: 28.9 -> 32.0 | Helped |
| add source-guided target-to-source translation and its unsupervised learning (ablation) | BiSIDA · GTA5-to-CityScapes | mIoU: 29.3 -> 31.8 | Helped |
| add source-guided target-to-source translation and its unsupervised learning (ablation) | BiSIDA · SYNTHIA-to-CityScapes | mIoU: 28.9 -> 31.4 | Helped |
| add source-guided target-to-source translation alongside target-guided translation (ablation) | BiSIDA · SYNTHIA-to-CityScapes | mIoU: 32.0 -> 40.2 | Helped |
| add pseudo-labeling to both style translations (ablation) | BiSIDA · GTA5-to-CityScapes | mIoU: 35.1 -> 35.4 | Helped |
| add self-ensembling to both style translations (ablation) | BiSIDA · GTA5-to-CityScapes | mIoU: 35.1 -> 39.4 | Helped |
| add both pseudo-labeling and self-ensembling (ablation) | BiSIDA · GTA5-to-CityScapes | mIoU: 39.4 -> 43.2 | Helped |
| add color perturbation without source-guided style translation (ablation) | BiSIDA unsupervised branch · SYNTHIA-to-CityScapes | mIoU: 32.2 -> 32.1 | Hurt |
| add source-guided translation alongside color perturbation (ablation) | BiSIDA unsupervised branch · SYNTHIA-to-CityScapes | mIoU: 32.1 -> 42.1 | Helped |
Closest papers by meaning
- PixMatch: Unsupervised Domain Adaptation via Pixelwise Consistency Training (CVPR 2021 · similarity 0.91)
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.91)
- Unsupervised Model Adaptation for Continual Semantic Segmentation (AAAI 2021 · similarity 0.91)
- Style Mixing and Patchwise Prototypical Matching for One-Shot Unsupervised Domain Adaptive Semantic Segmentation (AAAI 2022 · similarity 0.90)
- Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation (CVPR 2019 · similarity 0.90)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.90)
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.