Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Semantic Segmentation With Cross-Consistency Training
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 |
|---|---|---|---|
| Enforce prediction consistency across perturbed encoder representations using unlabeled examples | CCT · PASCAL VOC val, n=1000 | mIoU: CCT 58.6; baseline 51.4 | Helped |
| Add image-level labeled examples alongside pixel-level labels | CCT · PASCAL VOC val | mIoU: 3.8 points | Helped |
| Enforce consistency across domains with a shared encoder | CCT · Cityscapes + CamVid | average mIoU: 35.6 -> 44.4; 35.9 -> 47.9 | Helped |
| Enforce consistency across domains despite non-overlapping label spaces | CCT · Cityscapes + SUN RGB-D | average mIoU: 46.2 -> 52.1 | Helped |
| Apply input-level adversarial perturbations for consistency (ablation) | PASCAL VOC semi-supervised segmentation · PASCAL VOC val | mIoU: 51.4 -> 50.0; 59.2 -> 57.9 | Hurt |
| Use Mean Teachers consistency instead of cross-consistency on hidden representations (ablation) | PASCAL VOC semi-supervised segmentation · PASCAL VOC val | mIoU: 51.4 -> 51.3; 59.2 -> 59.4 | No clear effect |
| Replace MSE consistency distance with KL divergence (ablation) | CCT · PASCAL VOC val | mIoU: 58.6 -> 54.0; 64.4 -> 62.5 | Hurt |
| Replace MSE consistency distance with JS divergence (ablation) | CCT · PASCAL VOC val | mIoU: 58.6 -> 58.4; 64.4 -> 64.3 | No clear effect |
| Mask uncertain main predictions during unlabeled consistency training (ablation) | CCT · PASCAL VOC val | mIoU: 58.6 vs 58.4; 64.4 vs 63.3 | No clear effect |
| Add pairwise consistency loss between auxiliary predictions (ablation) | CCT · PASCAL VOC val | mIoU: 58.6 -> 55.6; 64.4 -> 61.2 | Hurt |
| Add adversarial domain-alignment loss (ablation) | CCT · Cityscapes + CamVid | average mIoU: 44.4 -> 42.2; 47.9 -> 45.2 | Hurt |
Closest papers by meaning
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.91)
- ScaleMatch: Multi-scale Consistency Enhancement for Semi-supervised Semantic Segmentation (AAAI 2025 · similarity 0.90)
- RankMatch: Exploring the Better Consistency Regularization for Semi-supervised Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Relaxed Class-consensus Consistency for Semi-supervised Semantic Segmentation (AAAI 2025 · similarity 0.89)
- Towards the Uncharted: Density-Descending Feature Perturbation for Semi-supervised Semantic Segmentation (CVPR 2024 · similarity 0.89)
- RWMS: Reliable Weighted Multi-Phase for Semi-supervised Segmentation (AAAI 2024 · 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.