Research map / Label-efficient and robust learning / Domain adaptation and generalization
Open-Vocabulary Domain Generalization in Urban-Scene Segmentation
Dong Zhao, Qi Zang, Nan Pu, Wenjing Li, Nicu Sebe, Zhun Zhong
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 |
|---|---|---|---|
| replace cross-attention aggregation with selective SSM (ablation) | CAT-Seg · Dv-19 avg, ViT-B/16 (CS-7) | mIoU: 43.5 -> 45.6 | Improved |
| replace cross-attention aggregation with selective SSM (ablation) | CAT-Seg · Dv-58 avg, ViT-B/16 (CS-7) | mIoU: 43.5 -> 44.1 | Improved |
| add image/text-conditioned modulation (ablation) | SSM aggregation · Dv-19 avg, ViT-B/16 (CS-7) | mIoU: 45.6 -> 47.6 | Improved |
| add learnable geometric decay (ablation) | SSM aggregation · Dv-19 avg, ViT-B/16 (CS-7) | mIoU: 47.6 -> 48.3 | Improved |
| add chunk-wise aggregation (ablation) | SSM aggregation · Dv-19 avg, ViT-L/14 (CS-7) | mIoU: 53.2 -> 55.3 | Improved |
| add snake scanning (ablation) | SSM aggregation · Dv-58 avg, ViT-B/16 (CS-7) | mIoU: 47.3 -> 47.9 | Improved |
| use S2-Corr instead of CAT-Seg (real-to-real) | CAT-Seg · Dv-19 avg, ViT-B/16 (CS-7) | mIoU: 43.5 -> 50.3 | Improved |
| use S2-Corr instead of CAT-Seg (synthetic-to-real) | CAT-Seg · Dv-58 avg, ViT-L/14 (GTA-7) | mIoU: 48.2 -> 49.4 | Improved |
| train on COCO-Stuff instead of driving data | CAT-Seg · Dv-19 | mIoU: 49.3 -> 31.6 | Hurt |
| move training domain from synthetic GTA to real Cityscapes | CAT-Seg · Dv-19 | mIoU: 47.5 -> 49.3 | Improved |
| use S2-Corr instead of CAT-Seg for inference efficiency | CAT-Seg · vocab150 throughput | FPS: 5.7 -> 18.3 | Improved |
| use S2-Corr instead of CAT-Seg for GPU memory | CAT-Seg · GPU memory | GPU memory (GB): 13.8 -> 9.2 | Improved |
Closest papers by meaning
- Leveraging Depth and Language for Open-Vocabulary Domain-Generalized Semantic Segmentation (NeurIPS 2025 · similarity 0.94)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · similarity 0.90)
- Stronger Fewer & Superior: Harnessing Vision Foundation Models for Domain Generalized Semantic Segmentation (CVPR 2024 · similarity 0.90)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
- Generalize or Detect? Towards Robust Semantic Segmentation Under Multiple Distribution Shifts (NeurIPS 2024 · similarity 0.89)
- Exploring Semantic Consistency and Style Diversity for Domain Generalized Semantic Segmentation (AAAI 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.
