Research map / Label-efficient and robust learning / Domain adaptation and generalization
Uncertainty-guided Open-Set Source-Free Unsupervised Domain Adaptation with Target-private Class Segregation
Mattia Litrico, Davide Talon, Sebastiano Battiato, Alessio Del Bue, Mario Valerio Giuffrida, Pietro Morerio
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 clustering-based target classifier initialisation (ablation) | SF-OSDA pseudo-labelling · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 35.3 -> 77.9 | Improved |
| add neighbours-consensus uncertainty sample selection (ablation) | SF-OSDA pseudo-labelling · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 77.9 -> 80.6 | Improved |
| add class-separation uncertainty estimation (ablation) | SF-OSDA pseudo-labelling · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 80.6 -> 82.4 | Improved |
| add NL-InfoNCELoss contrastive loss (ablation) | SF-OSDA pseudo-labelling · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 82.4 -> 85.4 | Improved |
| use exponential instead of linear F for neighbours consensus coefficient (ablation) | uncertainty weighting · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 82.2 -> 85.0 | Improved |
| use linear instead of exponential F for class separation coefficient (ablation) | uncertainty weighting · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 84.9 -> 85.0 | No clear effect |
| combine uncertainty samples with AND instead of OR (ablation) | uncertainty sample selection · Office31 (3 tasks D->W, A->W, D->A) | avg HOS: 83.1 -> 85.0 | Improved |
| segregate private samples into multiple unknown classes instead of a single one (ablation) | SF-OSDA method · Office31 (3 tasks D->W, A->W, D->A) | HOS | Improved |
| source-free method vs SHOT on Office31 | SF-OSDA method · Office31 SF open-set (6 tasks avg) | HOS: 67.0 -> 85.4 | Improved |
| source-free method vs USD on Office-Home | SF-OSDA method · Office-Home SF open-set (12 tasks avg) | HOS: 65.9 -> 66.0 | Improved |
Closest papers by meaning
- Dynamic Target Distribution Estimation for Source-Free Open-Set Domain Adaptation (AAAI 2025 · similarity 0.91)
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.91)
- Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias (CVPR 2024 · similarity 0.90)
- Unveiling the Unknown: Unleashing the Power of Unknown to Known in Open-Set Source-Free Domain Adaptation (CVPR 2024 · similarity 0.90)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.90)
- Low Category Uncertainty and High Training Potential Instance Learning for Unsupervised Domain Adaptation (AAAI 2024 · similarity 0.90)
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.
