Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Adaptive Semantic Segmentation With Self-Supervised Depth Estimation
Qin Wang, Dengxin Dai, Lukas Hoyer, Luc Van Gool, Olga Fink
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 self-supervised depth auxiliary task naively (ablation) | DACS · SYNTHIA-to-Cityscapes | mIoU: 48.3 -> 49.6 | Improved |
| add task feature correlation module to depth auxiliary (ablation) | SimpleAux · SYNTHIA-to-Cityscapes | mIoU: 49.6 -> 54.2 | Improved |
| add task feature correlation module to depth auxiliary (ablation) | SimpleAux · SYNTHIA-to-Cityscapes (13 classes) | mIoU*: 55.9 -> 62.4 | Improved |
| add correlation learning modules (ablation) | CorDA without correlation modules · SYNTHIA-to-Cityscapes | mIoU: 51.7 -> 54.2 | Improved |
| add pseudo-label refinement with adaptation difficulty (ablation) | CorDA (F) · SYNTHIA-to-Cityscapes | mIoU: 54.2 -> 55.0 | Improved |
| add pseudo-label refinement with adaptation difficulty (ablation) | CorDA (F) · SYNTHIA-to-Cityscapes (13 classes) | mIoU*: 62.4 -> 62.8 | Improved |
| use stereo depth instead of Monodepth2 monocular depth for Cityscapes (ablation) | CorDA · SYNTHIA-to-Cityscapes | mIoU: 54.5 -> 55.0 | Improved |
| use ImageNet+COCO weights instead of ImageNet-only (ablation) | CorDA · SYNTHIA-to-Cityscapes | mIoU: 54.6 -> 55.0 | No clear effect |
| use ImageNet+COCO weights instead of ImageNet-only (ablation) | CorDA · GTA5-to-Cityscapes | mIoU: 56.4 -> 56.6 | No clear effect |
| add CorDA to DACS framework | DACS · SYNTHIA-to-Cityscapes | mIoU: 48.3 -> 55.0 | Improved |
| add CorDA to DACS framework | DACS · GTA5-to-Cityscapes | mIoU: 52.1 -> 56.6 | Improved |
| use feature-correlation learning instead of adversarial depth use | DADA · SYNTHIA-to-Cityscapes | mIoU: 42.6 -> 55.0 | 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.
- SPIGAN — SPIGAN: Privileged Adversarial Learning from Simulation (ICLR 2019) · better on 1 of 1 shared result
- DADA — DADA: Depth-Aware Domain Adaptation in Semantic Segmentation (ICCV 2019) · better on 1 of 1 shared result
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
- Improving Semi-Supervised and Domain-Adaptive Semantic Segmentation with Self-Supervised Depth Estimation (IJCV 2023 · similarity 0.94)
- Learning from Scale-Invariant Examples for Domain Adaptation in Semantic Segmentation (ECCV 2020 · similarity 0.92)
- Self-Supervised Augmentation Consistency for Adapting Semantic Segmentation (CVPR 2021 · similarity 0.92)
- Unsupervised Intra-Domain Adaptation for Semantic Segmentation Through Self-Supervision (CVPR 2020 · similarity 0.92)
- Geometric Unsupervised Domain Adaptation for Semantic Segmentation (ICCV 2021 · similarity 0.91)
- Constructing Self-Motivated Pyramid Curriculums for Cross-Domain Semantic Segmentation: A Non-Adversarial Approach (ICCV 2019 · 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.
