Research map / Label-efficient and robust learning / Semi-supervised learning
Space Engage: Collaborative Space Supervision for Contrastive-Based Semi-Supervised Semantic Segmentation
Changqi Wang, Haoyu Xie, Yuhui Yuan, Chong Fu, Xiangyu Yue
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 |
|---|---|---|---|
| add mix pseudo-labeling (agreement of logit and representation pseudo-labels) (ablation) | contrastive S4 baseline · PASCAL VOC 2012 classic, 183 labels | mIoU: 70.32 -> 71.92 | Improved |
| add cross pseudo-labeling between spaces (ablation) | contrastive S4 baseline · PASCAL VOC 2012 classic, 92 labels | mIoU: 67.11 -> 67.21 | No clear effect |
| add similarity-based indicator for representation learning (ablation) | contrastive S4 baseline · PASCAL VOC 2012 classic, 183 labels | mIoU: 70.32 -> 71.50 | Improved |
| combine mix pseudo-labeling with similarity indicator (ablation) | contrastive S4 baseline · PASCAL VOC 2012 classic, 183 labels | mIoU: 70.32 -> 72.74 | Improved |
| mix pseudo-labeling instead of cross pseudo-labeling (ablation) | CSS · PASCAL VOC 2012 classic, 183 labels | mIoU: 71.98 -> 72.74 | Improved |
| use representation-space pseudo-labels alone instead of logit-space (ablation) | contrastive S4 baseline · PASCAL VOC 2012 classic, 183 labels | mIoU: 70.32 -> 67.52 | Hurt |
| use mixed pseudo-labels from two spaces (ablation) | pseudo-label quality · PASCAL VOC 2012, 92 labels | pseudo-label mIoU: 78.87 -> 82.17 | Improved |
| use similarity-only indicator in both spaces instead of confidence (ablation) | contrastive S4 baseline · PASCAL VOC 2012 classic, 92 labels | mIoU: 67.11 -> 66.16 | Hurt |
| CSS versus PCR on VOC blender with 1323 labels | CSS (mix) · PASCAL VOC 2012 blender, 1323 labels | mIoU: 80.71 -> 79.54 | Hurt |
Closest papers by meaning
- Pixel Contrastive-Consistent Semi-Supervised Semantic Segmentation (ICCV 2021 · similarity 0.88)
- Semi-Supervised Semantic Segmentation With Pixel-Level Contrastive Learning From a Class-Wise Memory Bank (ICCV 2021 · similarity 0.88)
- PRCL: Probabilistic Representation Contrastive Learning for Semi-Supervised Semantic Segmentation (IJCV 2024 · similarity 0.88)
- Boosting Semi-Supervised Semantic Segmentation with Probabilistic Representations (AAAI 2023 · similarity 0.87)
- RankMatch: Exploring the Better Consistency Regularization for Semi-supervised Semantic Segmentation (CVPR 2024 · similarity 0.87)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.85)
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.
