Research map / Label-efficient and robust learning / Domain adaptation and generalization
Towards Fewer Annotations: Active Learning via Region Impurity and Prediction Uncertainty for Domain Adaptive Semantic Segmentation
Check 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 |
|---|---|---|---|
| Use RIPU region selection instead of LabOR at the 40-pixel budget | DeepLab-v2 for GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 63.5 -> 65.5 | Helped |
| Use RIPU region selection instead of LabOR at the 2.2% region budget | DeepLab-v2 for GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 66.6 -> 69.6 | Helped |
| Use impurity-based regions rather than random 2.2% region selection | DeepLab-v2 for GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 63.8 -> 68.5 | Helped |
| Combine region impurity with prediction uncertainty rather than impurity alone (ablation) | RIPU selection on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 68.1 -> 68.5 | Helped |
| Combine region impurity with prediction uncertainty rather than uncertainty alone (ablation) | RIPU selection on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 66.2 -> 68.5 | Helped |
| Add source-domain local prediction consistency loss (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 68.5 -> 69.0 | Helped |
| Add target-domain negative learning loss (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 68.5 -> 69.2 | Helped |
| Add both consistency and negative learning losses (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 69.2 -> 69.6 | Helped |
| Replace adaptive regions with fixed rectangles (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 69.6 -> 67.7 | Hurt |
| Replace adaptive regions with superpixels (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 69.6 -> 66.8 | Hurt |
| Set negative threshold τ to 0.20 instead of the default 0.05 (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 69.62 -> 68.13 | Hurt |
| Use an appropriate threshold up to 0.10 instead of 0.20 (ablation) | RIPU on GTAV → Cityscapes · GTAV → Cityscapes | mIoU: 69.62; 68.81; 69.54; 69.17; 68.94; 68.13 | Mixed |
Closest papers by meaning
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.90)
- Density Matters: Improved Core-Set for Active Domain Adaptive Segmentation (AAAI 2024 · similarity 0.90)
- Reinforced active learning for image segmentation (ICLR 2020 · similarity 0.89)
- Hyperbolic Active Learning for Semantic Segmentation under Domain Shift (ICML 2024 · similarity 0.88)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.87)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · similarity 0.87)
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.