Research map / Label-efficient and robust learning / Domain adaptation and generalization
Class-Balanced Pixel-Level Self-Labeling 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 |
|---|---|---|---|
| Add pixel-level self-labeling to rectify self-training pseudo labels (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 47.8 -> 55.7 | Helped |
| Use class-balanced sampling and unequal distribution alignment (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 51.8 -> 55.7 | Helped |
| Retain self-training pseudo labels alongside self-labeling (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 39.4 -> 55.7 | Helped |
| Initialize the self-labeling head with category prototypes instead of random weights (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 49.9 -> 55.7 | Helped |
| Add data augmentation (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 54.2 -> 55.7 | Helped |
| Add a momentum encoder (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 54.6 -> 55.7 | Helped |
| Add CPSL rather than ProDA | GTA5 → Cityscapes · Cityscapes validation | mIoU: 53.7 -> 55.7 | Helped |
| Add two knowledge distillation rounds to CPSL | CPSL · GTA5 → Cityscapes | mIoU: 55.7 -> 60.8 | Helped |
| Add CPSL rather than SAC | SYNTHIA → Cityscapes · Cityscapes validation | mIoU¹⁶: 52.6 -> 54.4 | Helped |
| Add CPSL rather than SAC | SYNTHIA → Cityscapes · Cityscapes validation | mIoU¹³: 59.3 -> 61.7 | Helped |
| Use 512 rather than 64, 128, or 256 sampled pixels per image (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 54.9, 55.3, 55.5 -> 55.7 | Helped |
| Increase samples per image from 512 to 1024 or 2048 (ablation) | CPSL · GTA5 → Cityscapes | mIoU: 55.7 -> 54.3; 55.7 -> 53.4 | Hurt |
Closest papers by meaning
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.92)
- Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021 · similarity 0.92)
- Unsupervised Intra-Domain Adaptation for Semantic Segmentation Through Self-Supervision (CVPR 2020 · similarity 0.90)
- Stable Neighbor Denoising for Source-free Domain Adaptive Segmentation (CVPR 2024 · similarity 0.88)
- Connectivity-Driven Pseudo-Labeling Makes Stronger Cross-Domain Segmenters (NeurIPS 2024 · similarity 0.87)
- Universal Domain Adaptation 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.