Research map / Label-efficient and robust learning / Domain adaptation and generalization
Exploring Robustness of Unsupervised Domain Adaptation in Semantic Segmentation
Jinyu Yang, Chunyuan Li, Weizhi An, Hehuan Ma, Yuzhi Guo, Yu Rong, Peilin Zhao, Junzhou Huang
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 |
|---|---|---|---|
| train with adversarial self-supervision (ASSUDA) on top of BDL | BDL · GTA5 to Cityscapes VGG16, eps=0.25 | mIoU: 14.6 -> 32.8 | Improved |
| add ASSUDA to BDL | BDL · GTA5 to Cityscapes VGG16, eps=0.5 | mIoU: 7.1 -> 19.1 | Improved |
| add ASSUDA to BDL | BDL · GTA5 to Cityscapes ResNet101, eps=0.25 | mIoU: 19.9 -> 33.1 | Improved |
| add ASSUDA to BDL | BDL · SYNTHIA to Cityscapes VGG16, eps=0.25 | mIoU: 13.6 -> 33.6 | Improved |
| add ASSUDA to BDL | BDL · SYNTHIA to Cityscapes ResNet101, eps=0.5 | mIoU: 11.0 -> 11.7 | No clear effect |
| attack BDL with FGSM eps=0.1 from surrogate PSPNet | BDL VGG16 · GTA5 to Cityscapes | mIoU: 41.3 -> 30.5 | Hurt |
| increase FGSM eps to 0.5 on BDL | BDL ResNet101 · GTA5 to Cityscapes | mIoU: 48.5 -> 6.50 | Hurt |
| use perturbation magnitude eps_m=1 instead of 0.1 (ablation) | ASSUDA · GTA5 to Cityscapes VGG16, eps=0.25 | mIoU: 21.2 -> 32.8 | Improved |
| use eps_m=1 instead of 0.1 (ablation) | ASSUDA · GTA5 to Cityscapes VGG16, eps=0.5 | mIoU: 8.3 -> 19.1 | Improved |
| use eps_m=1 instead of 0.1 (ablation) | ASSUDA · SYNTHIA to Cityscapes ResNet101, eps=0.1 | mIoU: 45.5 -> 48.5 | Improved |
| use eps_m=1 instead of 0.1 (ablation) | ASSUDA · SYNTHIA to Cityscapes VGG16, eps=0.5 | mIoU: 8.9 -> 18.5 | Improved |
| train on adversarial examples with VGG16 on mild perturbation | ASSUDA · SYNTHIA to Cityscapes VGG16, eps=0.1 | mIoU drop: -0.8 | Improved |
Closest papers by meaning
- Evaluating the Adversarial Robustness of Semantic Segmentation: Trying Harder Pays Off (ECCV 2024 · similarity 0.91)
- Multi-Target Adversarial Frameworks for Domain Adaptation in Semantic Segmentation (ICCV 2021 · similarity 0.91)
- Towards Reliable Evaluation and Fast Training of Robust Semantic Segmentation Models (ECCV 2024 · similarity 0.90)
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.90)
- On the Robustness of Semantic Segmentation Models to Adversarial Attacks (CVPR 2018 · similarity 0.90)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · 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.
