Research map / Label-efficient and robust learning / Domain adaptation and generalization
Classes Matter: A Fine-grained Adversarial Approach to Cross-domain Semantic Segmentation
Haoran Wang, Tong Shen, Wei Zhang, Ling-Yu Duan, Tao Mei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 fine-grained adversarial training (ablation) | DeepLab-v2 ResNet-101 · GTA5 -> Cityscapes | mIoU: 36.8 -> 46.9 | Improved |
| add self distillation (ablation) | FADA · GTA5 -> Cityscapes | mIoU: 46.9 -> 49.2 | Improved |
| add multi-scale testing (ablation) | FADA + self distillation · GTA5 -> Cityscapes | mIoU: 49.2 -> 50.1 | Improved |
| use soft labels instead of hard labels for class knowledge (ablation) | FADA · GTA5 -> Cityscapes | mIoU: 45.7 -> 46.9 | Improved |
| use soft labels instead of hard labels for class knowledge (ablation) | FADA · SYNTHIA -> Cityscapes | mIoU: 40.8 -> 41.5 | Improved |
| add confidence clipping (threshold 0.9) to soft labels (ablation) | FADA · GTA5 -> Cityscapes | mIoU: 45.7 -> 46.9 | Improved |
| lower confidence clipping threshold from 0.9 to 0.7 (ablation) | FADA · GTA5 -> Cityscapes | mIoU: 46.9 -> 46.2 | Hurt |
| add class-level fine-grained discriminator to a global feature alignment baseline (ablation) | DeepLab-v2 ResNet-101 · GTA5 -> Cityscapes | mIoU: 39.4 -> 45.7 | Improved |
| add FADA adaptation to source-only VGG-16 | DeepLab-v2 VGG-16 · GTA5 -> Cityscapes | mIoU: 28.3 -> 43.8 | Improved |
Closest papers by meaning
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.93)
- Category-Level Adversarial Adaptation for Semantic Segmentation using Purified Features (TPAMI 2021 · similarity 0.93)
- Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation (CVPR 2019 · similarity 0.92)
- Cross-Domain Grouping and Alignment for Domain Adaptive Semantic Segmentation (AAAI 2021 · similarity 0.91)
- Domain transfer through deep activation matching (ECCV 2018 · similarity 0.90)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · 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.
