Research map / Label-efficient and robust learning / Domain adaptation and generalization
Two-phase Pseudo Label Densification for Self-training based Domain Adaptation
Inkyu Shin, Sanghyun Woo, Fei Pan, In So Kweon
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 TPLD | CRST(MRKLD) · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 47.0 -> 51.2 | Improved |
| add TPLD | CRST(MRKLD) · GTA5->Cityscapes (rare classes, Deeplabv2-ResNet101) | R-mIoU: 30.3 -> 35.1 | Improved |
| add TPLD | CRST(MRKLD) · SYNTHIA->Cityscapes (Deeplabv3-ResNet101, 13 classes) | mIoU*: 48.1 -> 55.7 | Improved |
| add TPLD | CBST · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 45.9 -> 47.8 | Improved |
| add TPLD | CRST(LRENT) · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 45.9 -> 47.3 | Improved |
| add TPLD | CBST · SYNTHIA->Cityscapes (Deeplabv2-ResNet101) | mIoU: 42.6 -> 45.6 | Improved |
| add TPLD | CRST(LRENT) · SYNTHIA->Cityscapes (Deeplabv2-ResNet101) | mIoU: 42.7 -> 47.0 | Improved |
| lower selection threshold of CRST (p=0.4) (ablation) | CRST · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 47.0 -> 47.1 | No clear effect |
| lower selection threshold of CRST (p=0.6) (ablation) | CRST · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 47.0 -> 45.7 | Hurt |
| add voting stage before easy-hard classification (ablation) | TPLD · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 38.1 -> 51.2 | Improved |
| add easy-hard classification stage (ablation) | TPLD · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 49.5 -> 51.2 | Improved |
| order voting first then easy-hard classification (ablation) | TPLD · GTA5->Cityscapes (Deeplabv2-ResNet101) | mIoU: 49.1 -> 51.2 | Improved |
Closest papers by meaning
- Toward Improving Robustness and Accuracy in Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.89)
- Cycle Self-Training for Domain Adaptation (NeurIPS 2021 · similarity 0.88)
- Confidence Regularized Self-Training (ICCV 2019 · similarity 0.88)
- TransPL: VQ-Code Transition Matrices for Pseudo-Labeling of Time Series Unsupervised Domain Adaptation (ICML 2025 · similarity 0.86)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · similarity 0.86)
- Pseudo-Labeling for Unsupervised Domain Adaptation with Kernel GLMs (NeurIPS 2026 · 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.
