Research map / Label-efficient and robust learning / Semi-supervised learning
PRCL: Probabilistic Representation Contrastive Learning for Semi-Supervised Semantic Segmentation
Haoyu Xie, Changqi Wang, Jian Cun Zhao, Yang Liu, Jun Dan, Chong Fu, Baigui Sun
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 probabilistic representation (ablation) | contrastive teacher-student baseline · VOC 2012 classic (92 labels) | mIoU: 66.91 -> 68.49 | Improved |
| add probabilistic representation (ablation) | contrastive teacher-student baseline · VOC 2012 classic (732 labels) | mIoU: 74.99 -> 76.00 | Improved |
| add global distribution prototype to PR (ablation) | PR model · VOC 2012 classic (92 labels) | mIoU: 68.49 -> 69.52 | Improved |
| add virtual negatives to PR (ablation) | PR model · VOC 2012 classic (92 labels) | mIoU: 68.49 -> 69.07 | Improved |
| add both GDP and virtual negatives to PR (ablation) | PR model · VOC 2012 classic (92 labels) | mIoU: 68.49 -> 70.23 | Improved |
| use GDP update instead of EMA prototype update (ablation) | Baseline+ (EMA) vs PR+GDP · VOC 2012 classic (92 labels) | mIoU: 68.27 -> 69.52 | Improved |
| add EMA prototype update to vanilla representation (ablation) | Baseline · VOC 2012 classic (92 labels) | mIoU: 66.91 -> 68.27 | Improved |
| add virtual negatives instead of memory bank (ablation) | contrastive baseline with PR and GDP · VOC 2012 classic (92 labels) | mIoU: 66.89 -> 68.01 | Improved |
| add memory bank of size 65536 (ablation) | contrastive baseline with PR and GDP · VOC 2012 classic (92 labels) | mIoU: 66.89 -> 67.11 | No clear effect |
| add memory bank of size 30000 (ablation) | contrastive baseline with PR and GDP · VOC 2012 classic (92 labels) | mIoU: 66.89 -> 66.05 | Hurt |
| add probabilistic representation and PRCL components to baseline | Baseline · Cityscapes val | Silhouette score: 0.2708 -> 0.3014 | Improved |
| use PRCL instead of baseline | Baseline · VOC 2012 representation space | Silhouette score: 0.3188 -> 0.3836 | Improved |
Closest papers by meaning
- Boosting Semi-Supervised Semantic Segmentation with Probabilistic Representations (AAAI 2023 · similarity 0.95)
- Semi-Supervised Semantic Segmentation With Pixel-Level Contrastive Learning From a Class-Wise Memory Bank (ICCV 2021 · similarity 0.90)
- Looking Beyond Single Images for Contrastive Semantic Segmentation Learning (NeurIPS 2021 · similarity 0.89)
- Pixel Contrastive-Consistent Semi-Supervised Semantic Segmentation (ICCV 2021 · similarity 0.88)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.87)
- RankMatch: Exploring the Better Consistency Regularization for Semi-supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
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.
