Research map / Label-efficient and robust learning / Domain adaptation and generalization
Pulling Target to Source: A New Perspective on Domain Adaptive Semantic Segmentation
Haochen Wang, Yujun Shen, Jingjing Fei, Wei Li, Liwei Wu, Yuxi Wang, Zhaoxiang Zhang
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 |
|---|---|---|---|
| pull target to source: pseudo-target queries with source keys (ablation) | DAFormer · GTA5->Cityscapes (DAFormer) | mIoU: 68.3 -> 70.0 | Improved |
| use pseudo-target instead of real target features as queries (ablation) | T2S-DA · GTA5->Cityscapes (DAFormer) | mIoU: 68.6 -> 70.0 | Improved |
| use source cross-entropy rather than pseudo-target cross-entropy (ablation) | T2S-DA · GTA5->Cityscapes (DAFormer) | mIoU: 69.0 -> 70.0 | Improved |
| add vanilla source contrast (ablation) | DAFormer · GTA5->Cityscapes (DAFormer) | mIoU: 68.3 -> 68.5 | No clear effect |
| add class-balanced query sampling and domain-equalized negative sampling together (ablation) | T2S-DA · GTA5->Cityscapes (DAFormer) | mIoU: 68.7 -> 70.0 | Improved |
| add class-balanced query sampling alone (ablation) | T2S-DA · GTA5->Cityscapes (DAFormer) | mIoU: 68.7 -> 68.9 | No clear effect |
| add dynamic re-weighting on the pull loss (ablation) | T2S-DA · GTA5->Cityscapes (DAFormer) | mIoU: 68.9 -> 70.0 | Improved |
| add dynamic re-weighting on the target loss (ablation) | T2S-DA · GTA5->Cityscapes (DAFormer) | mIoU: 68.9 -> 68.8 | Hurt |
| use InfoNCE instead of MSE alignment (ablation) | T2S-DA · GTA5->Cityscapes, ResNet-101 | mIoU: 58.6 -> 60.4 | Improved |
| add MSE alignment (ablation) | T2S-DA · GTA5->Cityscapes, MiT-B5 | mIoU: 68.3 -> 69.4 | Improved |
| add pull loss with FDA translation (ablation) | DAFormer · GTA5->Cityscapes (DAFormer) | mIoU: 68.8 -> 70.0 | Improved |
| add pull loss with ColorJitter translation (ablation) | DAFormer · GTA5->Cityscapes (DAFormer) | mIoU: 68.1 -> 69.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DAFormer — DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation (CVPR 2022) · better on 3 of 3 shared results
- HRDA — HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation (ECCV 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bi-directional Contrastive Learning for Domain Adaptive Semantic Segmentation (ECCV 2022 · similarity 0.92)
- Feature Re-Representation and Reliable Pseudo Label Retraining for Cross-Domain Semantic Segmentation (TPAMI 2022 · similarity 0.92)
- Cross-Domain Grouping and Alignment for Domain Adaptive Semantic Segmentation (AAAI 2021 · similarity 0.92)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.91)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.91)
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.
