Research map / Label-efficient and robust learning / Domain adaptation and generalization
Content-Consistent Matching for Domain Adaptive Semantic Segmentation
Guangrui Li, Guoliang Kang, Wu Liu, Yunchao Wei, Yi Yang
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 semantic layout matching source selection (ablation) | Self-training · GTA5 -> Cityscapes | mIoU: 48.1 -> 48.8 | Improved |
| add semantic layout matching source selection (ablation) | Self-training · SYNTHIA -> Cityscapes | mIoU: 41.2 -> 41.9 | Improved |
| add SLM plus pixel-wise similarity matching (full CCM) (ablation) | Self-training · GTA5 -> Cityscapes | mIoU: 48.1 -> 49.9 | Improved |
| add SLM plus pixel-wise similarity matching (full CCM) (ablation) | Self-training · SYNTHIA -> Cityscapes | mIoU: 41.2 -> 45.2 | Improved |
| add pixel-wise similarity matching on top of SLM (ablation) | CCM-SLM · GTA5 -> Cityscapes | mIoU: 48.8 -> 49.9 | Improved |
| add pixel-wise similarity matching on top of SLM (ablation) | CCM-SLM · SYNTHIA -> Cityscapes | mIoU: 41.9 -> 45.2 | Improved |
| update positive source set actively during training instead of fixing it at start (ablation) | CCM-Fix · GTA5 -> Cityscapes | mIoU: 48.9 -> 49.9 | Improved |
| update positive source set actively during training instead of fixing it at start (ablation) | CCM-Fix · SYNTHIA -> Cityscapes | mIoU: 44.2 -> 45.2 | Improved |
| select source samples with SLM at optimal gamma_img versus no SLM (ablation) | Self-training · GTA5 -> Cityscapes | mIoU: +0.7% | Improved |
| vary number of pixel neighbors K from 5 to 15 (ablation) | CCM · GTA5 -> Cityscapes | mIoU: 49.8, 49.9, 49.7 | No clear effect |
| add entropy regularization (ablation) | CCM · GTA5 -> Cityscapes | mIoU | Improved |
Closest papers by meaning
- Bi-directional Contrastive Learning for Domain Adaptive Semantic Segmentation (ECCV 2022 · similarity 0.90)
- Addressing Domain Gap via Content Invariant Representation for Semantic Segmentation (AAAI 2021 · similarity 0.90)
- Learning From Synthetic Data: Addressing Domain Shift for Semantic Segmentation (CVPR 2018 · similarity 0.90)
- Consistency Regularization with High-dimensional Non-adversarial Source-guided Perturbation for Unsupervised Domain Adaptation in Segmentation (AAAI 2021 · similarity 0.90)
- Classes Matter: A Fine-grained Adversarial Approach to Cross-domain Semantic Segmentation (ECCV 2020 · similarity 0.89)
- FDA: Fourier Domain Adaptation for Semantic Segmentation (CVPR 2020 · similarity 0.89)
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.
