Research map / Label-efficient and robust learning / Domain adaptation and generalization
C-SFDA: A Curriculum Learning Aided Self-Training Framework for Efficient Source Free 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 C-SFDA selective curriculum self-training instead of AdaCon | SFDA · VisDA | per-class accuracy (Avg.): 86.8 -> 87.8 | Helped |
| Use C-SFDA instead of the strongest listed SFDA baseline | SFDA · Office-31 | Top-1 accuracy (Avg.): 90.1 -> 90.5 | Helped |
| Use C-SFDA instead of the strongest listed SFDA baseline | SFDA · Office-Home | Top-1 accuracy (Avg.): 72.9 -> 73.5 | Helped |
| Use C-SFDA instead of AdaCon | SFDA · DomainNet | Top-1 accuracy (Avg.): 67.8 -> 69.0 | Helped |
| Use C-SFDA instead of HCL | semantic segmentation · GTA5→Cityscapes | mIoU: 48.1 -> 48.3 | Helped |
| Use C-SFDA instead of HCL | semantic segmentation · SYNTHIA→Cityscapes (16 categories) | mIoU: 43.5 -> 44.6 | Helped |
| Use C-SFDA instead of AUGCO | online test-time adaptation · Cityscapes→Dark-Zurich | mIoU: 32.4 -> 33.2 | Helped |
| Select pseudo-labels by confidence only rather than confidence and uncertainty (ablation) | C-SFDA · Office-Home | accuracy: 69.2 -> 71.8 | Hurt |
| Add uncertainty-based selection to confidence-based selection (ablation) | C-SFDA · VisDA | accuracy: 85.2 -> 87.4 | Helped |
| Add DoC selection to confidence-and-uncertainty selection (ablation) | C-SFDA · Office-Home | accuracy: 71.8 -> 73.3 | Helped |
| Remove label balancing (ablation) | C-SFDA · VisDA | accuracy: 87.8 -> 86.5 | Hurt |
| Remove contrastive loss while retaining reliable-label CE and propagation loss (ablation) | C-SFDA · Office-Home | accuracy: 71.6 -> 73.5 | Hurt |
Closest papers by meaning
- When Source-Free Domain Adaptation Meets Learning with Noisy Labels (ICLR 2023 · similarity 0.90)
- Understanding and Improving Source-free Domain Adaptation from a Theoretical Perspective (CVPR 2024 · similarity 0.90)
- Source-Free Domain Adaptation for Semantic Segmentation (CVPR 2021 · similarity 0.90)
- Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021 · similarity 0.89)
- Universal Semi-Supervised Domain Adaptation by Mitigating Common-Class Bias (CVPR 2024 · similarity 0.88)
- Source-Free Domain Adaptation with Frozen Multimodal Foundation Model (CVPR 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.