Research map / Label-efficient and robust learning / Domain adaptation and generalization
CDTrans: Cross-domain Transformer for Unsupervised Domain Adaptation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use two-way center-aware pseudo labeling instead of two-way matching alone (ablation) | CDTrans · VisDA-2017 | average accuracy: 77.1 -> 88.4 | Helped |
| Use center-aware filtering to refine two-way pseudo-label pairs (ablation) | CDTrans · VisDA-2017 | pair precision: 81.8 -> 91.3 | Helped |
| Use two-way rather than one-way pair selection (ablation) | CDTrans · VisDA-2017 | average accuracy: 76.6 -> 77.1; 76.9 -> 77.1 | Helped |
| Add target-branch classification loss to source-branch classification (ablation) | CDTrans · VisDA-2017 | average accuracy: 67.1 -> 86.7 | Helped |
| Add source-target branch classification loss (ablation) | CDTrans · VisDA-2017 | average accuracy: 86.7 -> 87.7 | Helped |
| Replace source-target classification loss with distillation loss (ablation) | CDTrans · VisDA-2017 | average accuracy: 87.7 -> 88.4 | Helped |
| Use CDTrans instead of source-only DeiT-base | DeiT-base UDA · VisDA-2017 | average accuracy: 67.1 -> 88.4 | Helped |
| Use CDTrans instead of source-only DeiT-base | DeiT-base UDA · Office-Home | average accuracy: 74.8 -> 80.5 | Helped |
| Use CDTrans instead of source-only DeiT-base | DeiT-base UDA · Office-31 | average accuracy: 88.8 -> 92.6 | Helped |
| Use CDTrans instead of source-only DeiT-base | DeiT-base UDA · DomainNet | average accuracy: 36.2 -> 45.2 | Helped |
Closest papers by meaning
- Progressive Feature Alignment for Unsupervised Domain Adaptation (CVPR 2019 · similarity 0.90)
- Aligning Domain-Specific Distribution and Classifier for Cross-Domain Classification from Multiple Sources (AAAI 2019 · similarity 0.89)
- Learning Transferrable Representations for Unsupervised Domain Adaptation (NeurIPS 2016 · similarity 0.89)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.89)
- Prompt-Based Distribution Alignment for Unsupervised Domain Adaptation (AAAI 2024 · similarity 0.88)
- Enhancing Domain Adaptation through Prompt Gradient Alignment (NeurIPS 2024 · similarity 0.87)
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.