Research map / Label-efficient and robust learning / Domain adaptation and generalization
An Adversarial Perturbation Oriented Domain Adaptation Approach for 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 pointwise adversarial feature perturbations and adversarial defense | Semantic segmentation domain adaptation · GTA5 → Cityscapes (ResNet-101) | mIoU: 43.8 -> 45.9; +2.1% | Helped |
| use pointwise adversarial feature perturbations and adversarial defense | Semantic segmentation domain adaptation · SYNTHIA → Cityscapes (ResNet-101) | mIoU₁₃: 47.8 -> 53.1; +5.5% | Helped |
| add feature perturbation attacks (ablation) | I-FGSPM framework · SYNTHIA → Cityscapes | mIoU₁₃: 44.8 -> 53.1 | Helped |
| use sign-preposed I-FGSPM instead of I-FGSM (ablation) | Feature perturbation generation · SYNTHIA → Cityscapes | mIoU₁₃: 51.8 -> 53.1; +1.3% | Helped |
| use non-iterative FGSPM instead of iterative I-FGSPM (ablation) | Feature perturbation generation · SYNTHIA → Cityscapes | mIoU₁₃: 52.9 vs 53.1 | No clear effect |
| use momentum-based MI-FGSPM (ablation) | Feature perturbation generation · SYNTHIA → Cityscapes | mIoU₁₃: 53.1 -> 52.2 | Hurt |
| perturb hidden features after layer4 rather than pixels or earlier layers (ablation) | Feature perturbation generation · SYNTHIA → Cityscapes | mIoU₁₃: 50.4 -> 53.1 | Helped |
| add entropy minimization (ablation) | Source-strategy segmentation adaptation · GTA5 → Cityscapes | mIoU: 36.6 -> 41.8 | Helped |
| add Lovász loss without perturbations (ablation) | Source-strategy segmentation adaptation · GTA5 → Cityscapes | mIoU: 36.6 -> 35.0 | Hurt |
| add Lovász loss without perturbations (ablation) | Source-strategy segmentation adaptation · SYNTHIA → Cityscapes | mIoU₁₃: 38.6 -> 41.3 | Helped |
| add perturbations, Lovász loss, and entropy minimization (ablation) | Source-strategy segmentation adaptation · GTA5 → Cityscapes | mIoU: 36.6 -> 45.9 | Helped |
| replace source-strategy initialization with ASN while retaining perturbations, Lovász, and entropy (ablation) | Full method · GTA5 → Cityscapes | mIoU: 45.9 -> 45.2 | Hurt |
Closest papers by meaning
- Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation (CVPR 2019 · similarity 0.94)
- Joint Adversarial Learning for Domain Adaptation in Semantic Segmentation (AAAI 2020 · similarity 0.92)
- Consistency Regularization with High-dimensional Non-adversarial Source-guided Perturbation for Unsupervised Domain Adaptation in Segmentation (AAAI 2021 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.89)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.89)
- Open-Set Domain Adaptation for Semantic 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.