Research map / Label-efficient and robust learning / Domain adaptation and generalization
Reducing Domain Gap by Reducing Style Bias
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 |
|---|---|---|---|
| Use SagNet instead of DeepAll | ResNet-18 · PACS multi-source DG | accuracy (%): 79.26 -> 83.25 | Helped |
| Use SagNet instead of DeepAll | ResNet-18 · Office-Home multi-source DG | accuracy (%): 60.82 -> 62.34 | Helped |
| Use SagNet instead of the ResNet-18 baseline | ResNet-18 · PACS single-source DG | accuracy (%): 54.3 -> 61.9 | Helped |
| Add SagNet to CDAN | ResNet-50 CDAN · Office-Home UDA | accuracy (%): 64.0 -> 65.3 | Helped |
| Add SagNet to CDAN | ResNet-18 CDAN · DomainNet UDA | accuracy (%): 53.4 -> 54.5 | Helped |
| Add SagNet to MME | ResNet-34 · DomainNet SSDA, 1 labeled target example | accuracy (%): 66.4 -> 68.6 | Helped |
| Add content-biased learning (ablation) | SagNet with ResNet-18 · PACS multi-source DG | accuracy (%): 81.15 -> 83.25 | Helped |
| Add adversarial style-biased learning (ablation) | SagNet with ResNet-18 · PACS multi-source DG | accuracy (%): 82.37 -> 83.25 | Helped |
| Add content-biased learning (ablation) | SagNet with ResNet-18 · Office-Home multi-source DG | accuracy (%): 61.52 -> 62.34 | Helped |
| Add adversarial style-biased learning (ablation) | SagNet with ResNet-18 · Office-Home multi-source DG | accuracy (%): 61.29 -> 62.34 | Helped |
| Increase the adversarial coefficient beyond its suitable range (ablation) | SagNet · PACS multi-source DG | accuracy (%) | Hurt |
| Apply style and content randomization after stage 3 rather than at other levels (ablation) | SagNet · PACS multi-source DG | accuracy (%) | Mixed |
Closest papers by meaning
- Style Neophile: Constantly Seeking Novel Styles for Domain Generalization (CVPR 2022 · similarity 0.88)
- Style Blind Domain Generalized Semantic Segmentation via Covariance Alignment and Semantic Consistence Contrastive Learning (CVPR 2024 · similarity 0.86)
- ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression (NeurIPS 2025 · similarity 0.86)
- Towards domain-invariant Self-Supervised Learning with Batch Styles Standardization (ICLR 2024 · similarity 0.85)
- Exploring Semantic Consistency and Style Diversity for Domain Generalized Semantic Segmentation (AAAI 2025 · similarity 0.85)
- Style Adaptation and Uncertainty Estimation for Multi-Source Blended-Target Domain Adaptation (NeurIPS 2024 · similarity 0.85)
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.