Research map / Label-efficient and robust learning / Domain adaptation and generalization
Hierarchical Unsupervised Relation Distillation for Source Free Domain Adaptation
Bowei Xing, Xianghua Ying, Ruibin Wang, Ruohao Guo, Ji Shi, Wenzhen Yue
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 local clustering relation distillation (ablation) | SHOT baseline · VisDA-2017 | average per-class accuracy: 82.9 -> 87.8 | Improved |
| add global semantic relation distillation (ablation) | SHOT baseline · VisDA-2017 | average per-class accuracy: 82.9 -> 87.0 | Improved |
| combine local and global relations (ablation) | HRD · VisDA-2017 | average per-class accuracy: 87.8 -> 89.0 | Improved |
| add adaptive ensemble re-weighting (ablation) | HRD · Office-Home | average accuracy: 74.7 -> 75.2 | Improved |
| add adaptive ensemble re-weighting (ablation) | HRD · DomainNet | average accuracy: 68.9 -> 69.9 | Improved |
| add source-like/target-specific data division (ablation) | HRD · VisDA-2017 | average per-class accuracy: 87.3 -> 89.8 | Improved |
| add class-balanced division threshold (ablation) | HRD · DomainNet | average accuracy: 69.4 -> 69.9 | Improved |
| use relation distillation instead of logit distillation (ablation) | HRD · VisDA-2017 | accuracy: 85.2 -> 89.8 | Improved |
| use teacher-student logit distillation instead of hard pseudo labels (ablation) | SFDA self-training · VisDA-2017 | accuracy: 82.6 -> 85.2 | Improved |
| increase neighbors K beyond 5 (ablation) | HRD · VisDA-2017 | average accuracy | Hurt |
| vary EMA coefficient among 0.95, 0.99, 0.999 (ablation) | HRD · VisDA-2017 | average accuracy: above 89.0% | No clear effect |
| add HRD to SHOT baseline | SHOT · Office-Home | average accuracy: 71.8 -> 75.2 | Improved |
Closest papers by meaning
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.90)
- Generalized Source-Free Domain Adaptation (ICCV 2021 · similarity 0.90)
- Class Relationship Embedded Learning for Source-Free Unsupervised Domain Adaptation (CVPR 2023 · similarity 0.90)
- When Source-Free Domain Adaptation Meets Learning with Noisy Labels (ICLR 2023 · similarity 0.90)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.89)
- Gradient-Based Sample Selection for Black-Box Universal Domain Adaptation (AAAI 2025 · similarity 0.88)
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.
