Research map / Label-efficient and robust learning / Domain adaptation and generalization
DCAN: Dual Channel-wise Alignment Networks for Unsupervised Scene Adaptation
Zuxuan Wu, Xintong Han, Yen-Liang Lin, Mustafa Gokhan Uzunbas, Tom Goldstein, Ser Nam Lim, Larry S. Davis
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 channel-wise feature alignment in the segmentation network (ablation) | CycleGAN images · Synthia->Cityscapes | mIoU: 30.4 -> 31.7 | Improved |
| add channel-wise feature alignment in the segmentation network (ablation) | UNIT images · Synthia->Cityscapes | mIoU: 31.6 -> 32.7 | Improved |
| add channel-wise feature alignment in the segmentation network (ablation) | DCAN · Synthia->Cityscapes | mIoU: 33.8 -> 35.4 | Improved |
| train image generator and segmentation network end-to-end instead of two stages (ablation) | DCAN · Synthia->Cityscapes | mIoU: 33.7 -> 35.4 | Improved |
| use DCAN channel-wise image synthesis instead of CycleGAN (ablation) | FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 30.4 -> 33.8 | Improved |
| use channel-wise alignment instead of adversarial ADDA (ablation) | FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 34.0 -> 35.4 | Improved |
| align features after Conv3 rather than Conv7 (ablation) | DCAN FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 32.7 -> 35.4 | Improved |
| align features after Conv3 rather than Conv6 (ablation) | DCAN FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 33.2 -> 35.4 | Improved |
| align after Conv2 instead of no alignment (ablation) | DCAN FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 33.8 -> 34.0 | No clear effect |
| align at Conv7 instead of no alignment (ablation) | DCAN FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 33.8 -> 32.7 | Hurt |
| increase input resolution from 256x512 to 512x1024 (ablation) | DCAN FCN8s-VGG16 · Synthia->Cityscapes | mIoU: 29.6 -> 35.4 | Improved |
| add DCAN adaptation to source-only training | PSPNet · Gta5->Cityscapes | mIoU: 33.3 -> 41.7 | Improved |
Later papers that compared with this method
5 later accepted papers report DCAN as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Probability-Polarized Optimal Transport for Unsupervised Domain Adaptation (AAAI 2024 · better on 1 of 1 shared result)
- Semantic Concentration for Domain Adaptation (ICCV 2021 · better on 2 of 2 shared results)
- Gradient Distribution Alignment Certificates Better Adversarial Domain Adaptation (ICCV 2021 · better on 2 of 2 shared results)
- Adversarial Unsupervised Domain Adaptation With Conditional and Label Shift: Infer, Align and Iterate (ICCV 2021 · better on 1 of 1 shared result)
- Bidirectional Learning for Domain Adaptation of Semantic Segmentation (CVPR 2019 · better on 3 of 3 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
- Classes Matter: A Fine-grained Adversarial Approach to Cross-domain Semantic Segmentation (ECCV 2020 · similarity 0.89)
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.89)
- Domain transfer through deep activation matching (ECCV 2018 · similarity 0.89)
- Label-Driven Reconstruction for Domain Adaptation in Semantic Segmentation (ECCV 2020 · similarity 0.88)
- DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation (CVPR 2022 · similarity 0.88)
- Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation (CVPR 2019 · 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.
