Research map / Label-efficient and robust learning / Domain adaptation and generalization
SRasP: Self-Reorientation Adversarial Style Perturbation for Cross-Domain Few-Shot Learning
Wenqian Li, Pengfei Fang, Hui Xue
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 Self-Reorientation gradient ensemble to baseline (ablation) | baseline · CD-FSL 1-shot | average accuracy: 45.56 -> 46.94 | Improved |
| add Self-Reorientation gradient ensemble to baseline (ablation) | baseline · CD-FSL 5-shot | average accuracy: 62.07 -> 62.81 | Improved |
| add semantic consistency loss L_con to SR (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 46.94 -> 48.67 | Improved |
| add CDTO loss to SR (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 46.94 -> 47.15 | Improved |
| use full SR + CDTO + L_con (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 48.67 -> 50.24 | Improved |
| use incoherent crop mining instead of random crops (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 49.49 -> 50.24 | Improved |
| use incoherent crop mining instead of concept crops (ablation) | SRasP · CD-FSL 5-shot | average accuracy: 65.24 -> 65.78 | Improved |
| set reorientation factor xi=0.1 instead of 0.01 (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 49.49 -> 50.24 | Improved |
| set reorientation factor xi=1 instead of 0.1 (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 50.24 -> 49.58 | Hurt |
| set trade-off lambda=0.2 instead of 0 (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 49.69 -> 50.24 | Improved |
| increase lambda from 0.2 to 1 (ablation) | SRasP · CD-FSL 5-shot | average accuracy: 65.78 -> 64.50 | Hurt |
| use different kappa1 and kappa2 instead of the same (ablation) | SRasP · CD-FSL 1-shot | average accuracy: 49.30 -> 50.24 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SVasP — SVasP: Self-Versatility Adversarial Style Perturbation for Cross-Domain Few-Shot Learning (AAAI 2025) · better on 5 of 5 shared results
- StyleAdv — StyleAdv: Meta Style Adversarial Training for Cross-Domain Few-Shot Learning (CVPR 2023) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SVasP: Self-Versatility Adversarial Style Perturbation for Cross-Domain Few-Shot Learning (AAAI 2025 · similarity 0.95)
- StyleAdv: Meta Style Adversarial Training for Cross-Domain Few-Shot Learning (CVPR 2023 · similarity 0.89)
- Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.85)
- StyDeSty: Min-Max Stylization and Destylization for Single Domain Generalization (ICML 2024 · similarity 0.85)
- Generate Universal Adversarial Perturbations for Few-Shot Learning (NeurIPS 2024 · similarity 0.85)
- Flatten Long-Range Loss Landscapes for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.84)
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.
