Research map / Label-efficient and robust learning / Domain adaptation and generalization
CLIP-Powered Domain Generalization and Domain Adaptation: A Comprehensive Survey
Jindong Li, Yongguang Li, Yali Fu, Jiahong Liu, Yixin Liu, Menglin Yang, Irwin King
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 prompt tuning (CoOp) to zero-shot CLIP | CLIP ViT-B/16 · DG average (PACS, VLCS, Office-Home, DomainNet) | accuracy (%): 77.3 -> 80.2 | Improved |
| add image-conditional prompts (CoCoOp) over CoOp | CoOp · DG average (PACS, VLCS, Office-Home, DomainNet) | accuracy (%): 80.2 -> 80.2 | No clear effect |
| add style-aware prompt learning (StyLIP) over CoOp | CoOp · DG average (PACS, VLCS, Office-Home, DomainNet) | accuracy (%): 80.2 -> 82.9 | Improved |
| add deeper multimodal prompts (MaPLe) over CoOp | CoOp · DG average (PACS, VLCS, Office-Home, DomainNet) | accuracy (%): 80.2 -> 80.4 | No clear effect |
| use source-free style prompts (PromptStyler) instead of source-based CoOp | CoOp · DG average (PACS, VLCS, Office-Home, DomainNet) | accuracy (%): 80.2 -> 80.8 | No clear effect |
| add CLIPood over zero-shot CLIP | CLIP ViT-B/16 · DomainNet (DG) | accuracy (%): 54.1 -> 63.5 | Improved |
| add domain adaptive prompts (DAPL) to zero-shot CLIP | CLIP ResNet-50 · Office-Home DA average (ResNet-50) | accuracy (%): 71.8 -> 74.5 | Improved |
| add ProDe source-free adaptation over DIFO | DIFO · Office-Home DA average (ResNet-50) | accuracy (%): 79.4 -> 81.1 | Improved |
| scale CLIP backbone from ResNet-50 to ViT-B | CLIP zero-shot · Office-Home DA average | accuracy (%): 71.8 -> 82.4 | Improved |
| use ViT-B/32 instead of ViT-B/16 for ProDe | ProDe · Office-Home DA average | accuracy (%): 84.5 vs 81.1 ResNet; below ViT-B/16 DAMP 87.1 | Hurt |
| source-free CDBN vs source-available DAMP (ViT-B) | DAMP · Office-Home DA average (ViT-B) | accuracy (%): 87.1 -> 86.2 | Hurt |
| add CMKD over DAPL (ViT-B) | DAPL · Office-Home DA average (ViT-B) | accuracy (%): 84.4 -> 88.4 | Improved |
Closest papers by meaning
- Domain Generalization in CLIP via Learning with Diverse Text Prompts (CVPR 2025 · similarity 0.94)
- CLIPCEIL: Domain Generalization through CLIP via Channel rEfinement and Image-text aLignment (NeurIPS 2024 · similarity 0.94)
- OSLoPrompt: Bridging Low-Supervision Challenges and Open-Set Domain Generalization in CLIP (CVPR 2025 · similarity 0.91)
- Learning to Adapt Frozen CLIP for Few-Shot Test-Time Domain Adaptation (ICLR 2025 · similarity 0.90)
- Unknown Prompt the only Lacuna: Unveiling CLIP's Potential for Open Domain Generalization (CVPR 2024 · similarity 0.90)
- PracticalDG: Perturbation Distillation on Vision-Language Models for Hybrid Domain Generalization (CVPR 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.
