Research map / Label-efficient and robust learning / Domain adaptation and generalization
Training-free Cross-domain Few-shot Segmentation via Robust Semantic Representation and Matching
Sujun Sun, Mingwu Ren, Haofeng Zhang
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 SAFR feature re-fusion (ablation) | baseline DINOv3 matching · Average 1-shot | mIoU: 58.74 -> 62.53 | Improved |
| add SAFR feature re-fusion (ablation) | baseline DINOv3 matching · Average 5-shot | mIoU: 64.54 -> 68.64 | Improved |
| add ASE on top of SAFR (ablation) | SAFR · Average 1-shot | mIoU: 62.53 -> 66.16 | Improved |
| add HPM on top of SAFR (ablation) | SAFR · Average 1-shot | mIoU: 62.53 -> 65.99 | Improved |
| add HPM on top of SAFR+ASE (ablation) | SAFR+ASE · Average 1-shot | mIoU: 66.16 -> 68.39 | Improved |
| add HPM on top of SAFR+ASE (ablation) | SAFR+ASE · Average 5-shot | mIoU: 70.86 -> 73.08 | Improved |
| reduce residual weight of position-aware features instead of SAFR (ablation) | baseline DINOv3 · Average 1-shot | mIoU: 58.74 -> 58.28 | Hurt |
| variance regularization on position-sensitive channels (ablation) | baseline DINOv3 · Average 1-shot | mIoU: 58.74 -> 57.69 | Hurt |
| remove position-sensitive channels from F_last (ablation) | baseline DINOv3 · Average 1-shot | mIoU: 58.74 -> 62.15 | Improved |
| use SAFR instead of removing channels (ablation) | position-sensitive channel removal · Average 1-shot | mIoU: 62.15 -> 62.53 | Improved |
| add attention mask with fixed threshold 0.5 in ASE (ablation) | ASE w/o mask · Average 1-shot | mIoU: 63.76 -> 64.53 | Improved |
| use robust threshold instead of fixed threshold in ASE (ablation) | ASE · Average 1-shot | mIoU: 64.53 -> 66.16 | Improved |
Closest papers by meaning
- Adapt Before Comparison: A New Perspective on Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.93)
- Adapter Naturally Serves as Decoupler for Cross-Domain Few-Shot Semantic Segmentation (ICML 2025 · similarity 0.93)
- The Devil is in Low-Level Features for Cross-Domain Few-Shot Segmentation (CVPR 2025 · similarity 0.92)
- Cross-Domain Few-Shot Segmentation via Iterative Support-Query Correspondence Mining (CVPR 2024 · similarity 0.92)
- Dual-Agent Optimization framework for Cross-Domain Few-Shot Segmentation (CVPR 2025 · similarity 0.91)
- Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.91)
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.
