Research map / Label-efficient and robust learning / Domain adaptation and generalization
Category Contrast for Unsupervised Domain Adaptation in Visual Tasks
Check 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 |
|---|---|---|---|
| Add category contrast with a dictionary containing source- and target-domain keys | DeepLab-V2 with ResNet101 · GTA5 → Cityscapes | mIoU: 36.6 -> 49.2 | Helped |
| Add category contrast to ProDA | ProDA · GTA5 → Cityscapes | mIoU: 57.5 -> 58.0 | Helped |
| Use source-domain keys only rather than the full source-and-target dictionary (ablation) | CaCo · GTA5 → Cityscapes | mIoU: 49.2 -> 46.8 | Hurt |
| Use target-domain keys only rather than the full source-and-target dictionary (ablation) | CaCo · GTA5 → Cityscapes | mIoU: 49.2 -> 48.3 | Hurt |
| Add category contrast to CRST | CRST · SYNTHIA → Cityscapes | mIoU: 43.8 -> 48.5 | Helped |
| Add category contrast to CRDA | CRDA · Cityscapes → Foggy Cityscapes | mAP: 37.4 -> 40.2 | Helped |
| Add category contrast to CRDA | CRDA · Cityscapes → BDD100k | mAP: 26.9 -> 30.2 | Helped |
| Add category contrast to CRST | CRST · VisDA17 | Mean accuracy: 78.1 -> 81.6 | Helped |
| Add category contrast to CRST | CRST · Office-31 | Mean accuracy: 86.8 -> 88.1 | Helped |
| Vary dictionary queue length from 50 to 150 (ablation) | CaCo · GTA5 → Cityscapes | mIoU | No clear effect |
| Use a single shared temperature rather than key-specific temperatures (ablation) | CaCo · UDA experiments | performance | Hurt |
| Use two separate source and target dictionaries instead of one domain-mixed dictionary (ablation) | CaCo · UDA experiments | performance | Hurt |
Closest papers by meaning
- Connect, Not Collapse: Explaining Contrastive Learning for Unsupervised Domain Adaptation (ICML 2022 · similarity 0.92)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.88)
- Low Category Uncertainty and High Training Potential Instance Learning for Unsupervised Domain Adaptation (AAAI 2024 · similarity 0.88)
- Link-based Contrastive Learning for One-Shot Unsupervised Domain Adaptation (CVPR 2025 · similarity 0.88)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Domain-Agnostic Mutual Prompting for Unsupervised Domain Adaptation (CVPR 2024 · similarity 0.86)
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.