Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels
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 unreliable pixels as negative candidates rather than reliable pixels (ablation) | U²PL · PASCAL VOC 2012 blender val, 1/8 (1323) | mIoU: 77.30 -> 79.01 | Helped |
| Use unreliable pixels as negative candidates rather than sampling all pixels (ablation) | U²PL · Cityscapes val, 1/4 (744) | mIoU: 74.51 -> 76.47 | Helped |
| Add contrastive loss L_c (ablation) | U²PL supervised baseline · PASCAL VOC 2012 blender val, 1/4 (2646) | mIoU: 73.02 -> 77.08 | Helped |
| Add category-wise memory bank, PRT, and high-entropy filtering (ablation) | U²PL with contrastive loss · PASCAL VOC 2012 blender val, 1/4 (2646) | mIoU: 73.02 -> 78.49 | Helped |
| Add DPA and high-entropy filtering (ablation) | U²PL with contrastive loss · PASCAL VOC 2012 blender val, 1/4 (2646) | mIoU: 73.02 -> 79.07 | Helped |
| Add high-entropy filtering of unreliable pixels (ablation) | U²PL with other components · PASCAL VOC 2012 blender val, 1/4 (2646) | mIoU: 77.57 -> 79.30 | Helped |
| Set probability rank thresholds to r_l=3 and r_h=20 instead of r_l=3 and r_h=10 (ablation) | U²PL · PASCAL VOC 2012 blender val, 1/4 (2646) | mIoU: 78.91 -> 79.30 | Helped |
| Set alpha_0 to 20% rather than 40%, 30%, or 10% (ablation) | U²PL · PASCAL VOC 2012 blender val, 1/8 (1323) | mIoU: 76.77 / 77.34 / 77.80 -> 79.01 | Helped |
| Replace contrastive loss with binary cross-entropy loss (ablation) | U²PL · PASCAL VOC 2012 blender val, 1/4 (2646) | mIoU: 79.30 -> 79.64 | Helped |
| Replace contrastive loss with binary cross-entropy loss (ablation) | U²PL · Cityscapes val, 1/16, 1/8, 1/4, 1/2 | mIoU: 70.30 -> 69.87; 74.37 -> 72.93; 76.47 -> 75.91; 79.05 -> 78.36 | Mixed |
Closest papers by meaning
- PseudoSeg: Designing Pseudo Labels for Semantic Segmentation (ICLR 2021 · similarity 0.92)
- DAW: Exploring the Better Weighting Function for Semi-supervised Semantic Segmentation (NeurIPS 2023 · similarity 0.91)
- Semi-Supervised Semantic Segmentation With Cross-Consistency Training (CVPR 2020 · similarity 0.91)
- RWMS: Reliable Weighted Multi-Phase for Semi-supervised Segmentation (AAAI 2024 · similarity 0.90)
- Beyond Confidence: Exploiting Homogeneous Pattern for Semi-Supervised Semantic Segmentation (ICML 2025 · similarity 0.89)
- DuPL: Dual Student with Trustworthy Progressive Learning for Robust Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.88)
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.