Research map / Label-efficient and robust learning / Domain adaptation and generalization
FSDR: Frequency Space Domain Randomization for Domain Generalization
Check 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 |
|---|---|---|---|
| randomize domain-variant frequency components while preserving domain-invariant components (ablation) | FSDR semantic segmentation · GTA5→Cityscapes, Mapillary, and BDD | mIoU: 33.4 -> 44.8; 27.9 -> 43.4; 27.3 -> 41.2 | Helped |
| use spectrum-analysis randomization (ablation) | FSDR · GTA5→Cityscapes, Mapillary, and BDD | mIoU: 33.4 -> 42.1; 27.9 -> 39.2; 27.3 -> 37.8 | Helped |
| use iterative, adaptive spectrum-learning randomization (ablation) | FSDR · GTA5→Cityscapes, Mapillary, and BDD | mIoU: 33.4 -> 43.6; 27.9 -> 42.1; 27.3 -> 40.1 | Helped |
| combine spectrum analysis and spectrum learning (ablation) | FSDR · GTA5→Cityscapes, Mapillary, and BDD | mIoU: 42.1 -> 44.8; 39.2 -> 43.4; 37.8 -> 41.2 | Helped |
| use spectrum learning rather than spectrum analysis alone (ablation) | FSDR · GTA5→Cityscapes, Mapillary, and BDD | mIoU: 42.1 vs 43.6; 39.2 vs 42.1; 37.8 vs 40.1 | Helped |
| preserve a moderate fraction of FCs rather than all or none (ablation) | FSDR · GTA5→Cityscapes | mIoU: 33.4; 38.3 -> 41.0–44.8 | Helped |
| add FSDR losses to existing adaptation/generalization methods | Adapt-SegMap · GTA→Cityscapes, Mapillary, and BDD | mIoU: 41.4 -> 45.6; 38.3 -> 43.9; 36.2 -> 41.9 | Helped |
| add FSDR losses to existing adaptation/generalization methods | IBN-Net · GTA→Cityscapes, Mapillary, and BDD | mIoU: 40.3 -> 45.3; 35.9 -> 44.0; 35.6 -> 42.1 | Helped |
| remove high- and low-frequency components during source training (ablation) | spectrum-analysis classification model · SYNTHIA→ImageNet | target acc.: 65.2 -> 68.6; 65.2 -> 67.4 | Mixed |
Closest papers by meaning
- Learning Frequency-Adapted Vision Foundation Model for Domain Generalized Semantic Segmentation (NeurIPS 2024 · similarity 0.88)
- Dual-Agent Optimization framework for Cross-Domain Few-Shot Segmentation (CVPR 2025 · similarity 0.86)
- Lightweight Frequency Masker for Cross-Domain Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.85)
- Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.85)
- Exploring Semantic Consistency and Style Diversity for Domain Generalized Semantic Segmentation (AAAI 2025 · similarity 0.85)
- Gradient-Guided Annealing for Domain Generalization (CVPR 2025 · similarity 0.85)
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.