Research map / Label-efficient and robust learning / Semi-supervised learning
Augmentation Matters: A Simple-Yet-Effective Approach to 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 |
|---|---|---|---|
| Use AugSeg instead of the strongest listed baseline | ResNet-101 semi-supervised segmentation · Pascal VOC 2012 classic val, 1/8 (183 labels) | mIoU: 69.58 -> 75.45 | Helped |
| Use AugSeg instead of the strongest listed baseline | ResNet-101 semi-supervised segmentation · Cityscapes val, 1/4 (744 labels) | mIoU: 77.60 -> 79.56 | Helped |
| Use AugSeg instead of U²PL‡ | ResNet-101 semi-supervised segmentation · PASCAL VOCAug val, 1/8 (1323 labels), U²PL split | mIoU: 79.01 -> 81.46 | Helped |
| Add standard mean-teacher training (ablation) | supervised segmentation · Pascal VOC 2012 val, 366 labels | mIoU: 61.65 -> 69.06 | Helped |
| Add random intensity-based augmentation (ablation) | mean-teacher segmentation · Pascal VOC 2012 val, 366 labels | mIoU: 69.06 -> 72.41 | Helped |
| Add adaptive label-injecting augmentation (ablation) | mean-teacher segmentation · Pascal VOC 2012 val, 366 labels | mIoU: 69.06 -> 74.33 | Helped |
| Combine random intensity augmentation and adaptive label injection (ablation) | mean-teacher segmentation · Pascal VOC 2012 val, 366 labels | mIoU: 72.41 -> 76.17 | Helped |
| Add random intensity-based augmentation (ablation) | mean-teacher segmentation · Cityscapes val, 744 labels | mIoU: 75.96 -> 77.29 | Helped |
| Add adaptive label-injecting augmentation (ablation) | mean-teacher segmentation · Cityscapes val, 744 labels | mIoU: 75.96 -> 77.44 | Helped |
| Combine random intensity augmentation and adaptive label injection (ablation) | mean-teacher segmentation · Cityscapes val, 744 labels | mIoU: 77.44 -> 78.76 | Helped |
| Set unsupervised loss weight λu to 0.0 rather than 1.0 (ablation) | AugSeg · Pascal VOC 2012 val, 366 labels | mIoU: 76.17 -> 61.65 | Hurt |
| Set unsupervised loss weight λu to 2.0 rather than 1.0 (ablation) | AugSeg · Pascal VOC 2012 val, 366 labels | mIoU: 76.17 -> 77.05 | Helped |
Closest papers by meaning
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.91)
- Instance-Specific and Model-Adaptive Supervision for Semi-Supervised Semantic Segmentation (CVPR 2023 · similarity 0.91)
- Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant (NeurIPS 2022 · similarity 0.90)
- ST++: Make Self-Training Work Better for Semi-Supervised Semantic Segmentation (CVPR 2022 · similarity 0.90)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · similarity 0.88)
- AllSpark: Reborn Labeled Features from Unlabeled in Transformer for Semi-Supervised Semantic Segmentation (CVPR 2024 · 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.