Research map / Label-efficient and robust learning / Semi-supervised learning
Switching Temporary Teachers for Semi-Supervised Semantic Segmentation
Read the paperCheck 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 |
|---|---|---|---|
| Switch from one permanent teacher to two temporary teachers that alternate by epoch (ablation) | Dual Teacher · PASCAL VOC 2012 blended 1/16 | mIoU: 74.21 -> 77.00 | Helped |
| Increase temporary teachers from two to three while using two augmentations (ablation) | Dual Teacher · PASCAL VOC 2012 blended 1/16 | mIoU: 77.00 -> 76.56 | Hurt |
| Add a third augmentation (MixUp) to the two-augmentation pool (ablation) | Dual Teacher · PASCAL VOC 2012 blended 1/16 | mIoU: 77.00 -> 76.25 | Hurt |
| Ensemble teacher outputs instead of switching teachers by epoch (ablation) | Dual Teacher · PASCAL VOC 2012 blended, 1/16–1/2 partitions | mIoU: 77.00 -> 73.25; 79.23 -> 75.99; 78.94 -> 76.41; 80.18 -> 77.34 | Hurt |
| Add implicit consistency learning with uniform decay and drop rate 0.1 (ablation) | Dual Teacher · Cityscapes 1/16 | mIoU: 75.85 -> 76.81 | Helped |
| Increase uniform-decay drop rate above 0.1 (ablation) | Dual Teacher · Cityscapes 1/16 | mIoU: 76.81 -> 72.02; 72.30; 71.69 | Hurt |
| Replace original CutMix with adaptive CutMix (ablation) | Dual Teacher · PASCAL VOC 2012 blended, 1/16–1/4 partitions | mIoU: 77.00 -> 78.66; 79.23 -> 79.59; 78.94 -> 78.52 | Mixed |
| Use two augmentation types instead of one (ablation) | Dual Teacher · PASCAL VOC 2012 blended 1/16 | mIoU: 75.34 -> 77.00 | Helped |
| Use three augmentation types instead of two (ablation) | Dual Teacher · PASCAL VOC 2012 blended 1/16 | mIoU: 77.00 -> 76.25 | Hurt |
| Use Dual Teacher rather than Single Teacher | GTA5→Cityscapes unsupervised domain adaptation · GTA5→Cityscapes | mIoU: 49.71 -> 53.03 | Helped |
Closest papers by meaning
- Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant (NeurIPS 2022 · similarity 0.92)
- Learning from Future: A Novel Self-Training Framework for Semantic Segmentation (NeurIPS 2022 · similarity 0.90)
- Perturbed and Strict Mean Teachers for Semi-Supervised Semantic Segmentation (CVPR 2022 · similarity 0.89)
- DuPL: Dual Student with Trustworthy Progressive Learning for Robust Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.87)
- Relaxed Class-consensus Consistency for Semi-supervised Semantic Segmentation (AAAI 2025 · similarity 0.87)
- Distilling Knowledge from Heterogeneous Architectures for Semantic Segmentation (AAAI 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.