Research map / Label-efficient and robust learning / Domain adaptation and generalization
Bidirectional Learning for Domain Adaptation of Semantic Segmentation
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 target-domain adversarial segmentation adaptation (ablation) | Source-only segmentation model · GTA5 → Cityscapes | mIoU: 33.6 -> 40.9 | Helped |
| Translate source images into the target visual domain before segmentation training (ablation) | Source-only segmentation model · GTA5 → Cityscapes | mIoU: 33.6 -> 41.1 | Helped |
| Combine translation and adversarial segmentation adaptation in one bidirectional iteration (ablation) | GTA5 → Cityscapes adaptation · GTA5 → Cityscapes | mIoU: 41.1 -> 42.7 | Helped |
| Continue bidirectional training using the updated segmentation model (ablation) | GTA5 → Cityscapes adaptation · GTA5 → Cityscapes | mIoU: 42.7 -> 43.3 | Helped |
| Add two rounds of high-confidence target pseudo-label learning (ablation) | Bidirectional model M₀⁽¹⁾(F⁽¹⁾) · GTA5 → Cityscapes | mIoU: +4.5% | Helped |
| Add self-supervised learning in the second bidirectional iteration (ablation) | M₀⁽²⁾(F⁽²⁾) · GTA5 → Cityscapes | mIoU: 44.3 -> 48.5 | Helped |
| Use a 0.9 confidence threshold rather than a lower threshold for pseudo-label selection (ablation) | Self-supervised segmentation adaptation · GTA5 → Cityscapes | mIoU | Helped |
| Replace hard pseudo-label thresholding with soft confidence weighting (ablation) | Self-supervised segmentation adaptation · GTA5 → Cityscapes | mIoU | Hurt |
| Use more than one self-supervised update, up to convergence (ablation) | Self-supervised segmentation adaptation · GTA5 → Cityscapes | mIoU | Helped |
| Use bidirectional translation-segmentation learning with self-supervised learning | GTA5 → Cityscapes, ResNet101 · GTA5 → Cityscapes | mIoU: 48.5 vs 43.2 | Helped |
| Use bidirectional translation-segmentation learning with self-supervised learning | SYNTHIA → Cityscapes, ResNet101 · SYNTHIA → Cityscapes (13 classes) | mIoU: 51.4 vs 47.8 | Helped |
| Use bidirectional translation-segmentation learning with self-supervised learning | SYNTHIA → Cityscapes, VGG16 · SYNTHIA → Cityscapes (16 classes) | mIoU: 39.0 vs 35.4 | Helped |
Closest papers by meaning
- On the Road to Online Adaptation for Semantic Image Segmentation (CVPR 2022 · similarity 0.91)
- Self-Supervised Augmentation Consistency for Adapting Semantic Segmentation (CVPR 2021 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.89)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.89)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · similarity 0.88)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · similarity 0.88)
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.