Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unsupervised Feature Representation Learning for Domain-generalized Cross-domain Image Retrieval
Conghui Hu, Can Zhang, Gim Hee Lee
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 low-frequency phase augmentation (stage one) (ablation) | DG-UCDIR with L_rgb · DomainNet (6 train pairs, average) | P@50: 45.92 -> 50.15 | Improved |
| add amplitude augmentation (stage two) (ablation) | DG-UCDIR (v2) · DomainNet (6 train pairs, average) | P@50: 50.15 -> 53.77 | Improved |
| add phase-enhanced instance-instance contrastive loss (ablation) | DG-UCDIR (v3) · DomainNet (6 train pairs, average) | P@50: 53.77 -> 57.45 | Improved |
| add phase-enhanced instance-centroid contrastive loss (ablation) | DG-UCDIR (v4) · DomainNet (6 train pairs, average) | P@50: 57.45 -> 59.69 | Improved |
| use full proposed framework instead of strongest baseline BrAD | DG-UCDIR · PACS (average) | P@50: 45.33 -> 55.42 | Improved |
| use full proposed framework instead of strongest baseline BrAD | DG-UCDIR · DomainNet (average) | P@50: 50.91 -> 59.69 | Improved |
| choose different seen-domain training pairs for the same test domains (ablation) | DG-UCDIR · DomainNet Infograph-Sketch retrieval | P@50: 56.87 (Paint.+Clipart) vs 49.42 (Quick.+Real) | Mixed |
| generalize to unseen categories (cross-category) instead of seen categories | DG-UCDIR · PACS Photo-Sketch retrieval | P@50: 56.88 -> 34.79 | Hurt |
| use RANSAC-free: replace multi-domain unsupervised methods (CDS, PCS) with BrAD-style domain-gap handling | baselines · PACS | P@50: CDS 32.35, PCS 36.37 -> BrAD 45.33 | Improved |
Closest papers by meaning
- Unsupervised Cross-Domain Image Retrieval via Prototypical Optimal Transport (AAAI 2024 · similarity 0.91)
- Semantic Feature Learning for Universal Unsupervised Cross-Domain Retrieval (NeurIPS 2024 · similarity 0.91)
- Retrieval Across Any Domains via Large-scale Pre-trained Model (ICML 2024 · similarity 0.90)
- DiDA: Disambiguated Domain Alignment for Cross-Domain Retrieval with Partial Labels (AAAI 2024 · similarity 0.89)
- RoDA: Robust Domain Alignment for Cross-Domain Retrieval Against Label Noise (AAAI 2025 · similarity 0.88)
- Effective Comparative Prototype Hashing for Unsupervised Domain Adaptation (AAAI 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.
