Research map / Label-efficient and robust learning / Domain adaptation and generalization
Discover, Hallucinate, and Adapt: Open Compound Domain Adaptation for Semantic Segmentation
Read the paperCheck 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 Discover, Hallucinate, and domain-wise Adapt together | GTA5 to C-Driving segmentation · GTA5 to C-Driving | mIoU, C+O: 25.0 -> 32.3 | Helped |
| use the longer training scheme | DHA · GTA5 to C-Driving | mIoU, C+O: 29.2 -> 32.3 | Helped |
| cluster compound target images and use domain-specific adaptation (ablation) | Traditional UDA on GTA5 to C-Driving · GTA5 to C-Driving | mIoU, C+O: 29.3 -> 31.1 | Helped |
| add translated target-like source images (ablation) | Source Only · GTA5 to C-Driving | mIoU, C+O: 26.4 -> 30.4 | Helped |
| add traditional UDA after hallucinating target-like source images (ablation) | DHA hallucination variant · GTA5 to C-Driving | mIoU, C+O: 30.4 -> 31.0 | Helped |
| replace traditional UDA with domain-wise adversaries (ablation) | DHA · GTA5 to C-Driving | mIoU, C+O: 31.0 -> 32.3 | Helped |
| vary the number of latent target domains (ablation) | DHA · GTA5 to C-Driving | mIoU, C+O: 29.7 (K=2); 30.4 (K=3); 29.5 (K=4); 29.5 (K=5) | Mixed |
| add style-consistency loss during hallucination (ablation) | TGCF-DA · GTA5 to C-Driving | mIoU, C+O: 28.5 -> 30.4 | Helped |
| use DHA instead of the strongest listed UDA baseline | Synscapes to C-Driving · Synscapes to C-Driving | mIoU, C+O: 28.3 -> 29.6 | Helped |
| use DHA instead of AdaptSeg | SYNTHIA to C-Driving · SYNTHIA to C-Driving | mIoU, C+O: 20.2 -> 21.8 | Helped |
Closest papers by meaning
- Cluster, Split, Fuse, and Update: Meta-Learning for Open Compound Domain Adaptive Semantic Segmentation (CVPR 2021 · similarity 0.92)
- Learning to Adapt via Latent Domains for Adaptive Semantic Segmentation (NeurIPS 2021 · similarity 0.92)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.91)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · similarity 0.91)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.90)
- Towards Unsupervised Domain Bridging via Image Degradation in Semantic Segmentation (NeurIPS 2025 · 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.