Research map / Label-efficient and robust learning / Domain adaptation and generalization
Source Data-absent Unsupervised Domain Adaptation through Hypothesis Transfer and Labeling Transfer
Jian Liang, Dapeng Hu, Yunbo Wang, Ran He, Jiashi Feng
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 diversity loss L_div to information maximization (ablation) | SHOT-IM · VisDA-C (ResNet-50) | accuracy: 64.6 -> 73.9 | Improved |
| add self-supervised pseudo-labeling (ablation) | SHOT · VisDA-C (ResNet-50) | accuracy: 74.6 -> 76.7 | Improved |
| add relative rotation prediction (ablation) | SHOT · VisDA-C (ResNet-50) | accuracy: 74.8 -> 76.7 | Improved |
| add both self-supervisions to SHOT-IM (ablation) | SHOT-IM · VisDA-C (ResNet-50) | accuracy: 73.9 -> 76.7 | Improved |
| add labeling transfer (MixMatch) | SHOT · VisDA-C (ResNet-50) | accuracy: 76.7 -> 77.1 | Improved |
| add labeling transfer (MixMatch) | SHOT · Office-Home closed-set | accuracy: 72.0 -> 73.0 | Improved |
| add labeling transfer | Source-model-only · VisDA-C (ResNet-101) | accuracy: 48.0 -> 61.6 | Improved |
| add labeling transfer | SHOT on class truck · VisDA-C (ResNet-101) | per-class accuracy: 49.1 -> 28.8 | Hurt |
| add self-supervision (SHOT vs SHOT-IM) | SHOT-IM · Digits | accuracy: 97.9 -> 98.1 | Improved |
| add self-supervision (SHOT vs SHOT-IM) with beta=0 | SHOT-IM · Office-Home partial-set | accuracy: 76.7 -> 79.5 | Improved |
| add labeling transfer | SHOT · Office-Home semi-supervised (one-shot) | accuracy: 65.3 -> 66.1 | Improved |
| adapt with information maximization instead of source model only | Source-model-only · Office-Home closed-set | accuracy: 60.2 -> 70.5 | Improved |
Closest papers by meaning
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation (ICML 2020 · similarity 0.96)
- Unsupervised Multi-Source Domain Adaptation Without Access to Source Data (CVPR 2021 · similarity 0.91)
- Prior Knowledge Guided Unsupervised Domain Adaptation (ECCV 2022 · similarity 0.90)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.90)
- Guiding Pseudo-Labels With Uncertainty Estimation for Source-Free Unsupervised Domain Adaptation (CVPR 2023 · similarity 0.89)
- Link-based Contrastive Learning for One-Shot Unsupervised Domain Adaptation (CVPR 2025 · similarity 0.89)
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.
