Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Adaptive and Generalizable Network Architectures and Training Strategies for Semantic Image Segmentation
Lukas Hoyer, Dengxin Dai, Luc Van Gool
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 learning rate warmup (ablation) | DeepLabV2 DG · GTA->Cityscapes DG | mIoU: 25.2 -> 34.3 | Improved |
| add learning rate warmup (ablation) | DeepLabV2 UDA · GTA->Cityscapes UDA | mIoU: 49.1 -> 54.2 | Improved |
| replace ResNet with Transformer MiT encoder (ablation) | SegFormer-style UDA · GTA->Cityscapes UDA | mIoU: 54.2 -> 58.2 | Improved |
| add ImageNet thing-class feature distance (FD) (ablation) | MiT UDA · GTA->Cityscapes UDA | mIoU: 58.2 -> 61.7 | Improved |
| add rare class sampling (RCS) (ablation) | MiT UDA · GTA->Cityscapes UDA | mIoU: 58.2 -> 64.0 | Improved |
| use RCS instead of class-balanced sampling (T=infinity) (ablation) | MiT UDA · GTA->Cityscapes UDA | mIoU: 62.0 -> 64.0 | Improved |
| apply FD only to thing-classes (ablation) | MiT UDA · GTA->Cityscapes | mIoU: +1.4/+2.9 mIoU | Improved |
| stabilize self-training (ignore border pixels, alpha 0.999) (ablation) | DAFormer UDA · GTA->Cityscapes UDA | mIoU: 66.2 -> 67.0 | Improved |
| replace SegFormer decoder with DAFormer context-aware fusion decoder (ablation) | MiT UDA · GTA->Cityscapes UDA | mIoU: 67.0 -> 68.3 | Improved |
| restrict context fusion to bottleneck features only (ablation) | DAFormer decoder · GTA->Cityscapes UDA | mIoU: 68.3 -> 67.0 | Hurt |
| add multi-resolution training (HRDA) (ablation) | DAFormer · GTA->Cityscapes UDA | mIoU: 68.3 -> 73.8 | Improved |
| use learned scale attention instead of averaging scales (ablation) | HRDA · GTA->Cityscapes UDA | mIoU: 67.5 -> 71.5 | Improved |
Closest papers by meaning
- DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation (CVPR 2022 · similarity 0.96)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.91)
- Towards Unsupervised Domain Bridging via Image Degradation in Semantic Segmentation (NeurIPS 2025 · similarity 0.91)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance 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.
