Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Semantic Segmentation via Adaptive Equalization Learning
Hanzhe Hu, Fangyun Wei, Han Hu, Qiwei Ye, Jinshi Cui, Liwei Wang
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 |
|---|---|---|---|
| Use AEL instead of Cutmix-Seg | DeepLabv3+ with ResNet-101 · Cityscapes val 1/32 | mIoU: 69.11 -> 74.28 | Improved |
| Use AEL instead of Cutmix-Seg | DeepLabv3+ with ResNet-101 · PASCAL VOC 2012 val 1/32 | mIoU: 73.39 -> 76.97 | Improved |
| Use AEL instead of Cutmix-Seg | DeepLabv3+ with ResNet-101 · ADE20K val 1/32 | mIoU: 26.15 -> 28.40 | Improved |
| Add dynamic re-weighting to the basic framework (ablation) | AEL · Cityscapes val 1/32 and 1/16 | mIoU: +1.1% and +1.7% | Improved |
| Add adaptive equalization sampling to the basic framework (ablation) | AEL · Cityscapes val 1/32 and 1/16 | mIoU: +2.5% and +2.0% | Improved |
| Add adaptive CutMix to the basic framework (ablation) | AEL · Cityscapes val 1/32 and 1/16 | mIoU: +1.3% / +1.7% | Improved |
| Add adaptive Copy-Paste to the basic framework (ablation) | AEL · Cityscapes val 1/32 and 1/16 | mIoU: +0.5% / +0.5% | Improved |
| Use confidence rather than entropy or margin as the AES indicator (ablation) | AEL with AES · Cityscapes val 1/32 and 1/16 | mIoU: Ent: 71.38, 73.21; Conf: 72.51, 74.39; Margin: 70.86, 73.05 | Mixed |
| Set dynamic re-weighting gamma to 2 rather than 0 (ablation) | AEL with dynamic re-weighting · Cityscapes val 1/32 and 1/16 | mIoU: 69.11 -> 70.27; 72.13 -> 73.85 | Improved |
| Set adaptive CutMix ratio threshold to 0.005 (ablation) | AEL with adaptive CutMix · Cityscapes val 1/32 and 1/16 | mIoU: 73.43; 75.12 | Mixed |
| Set sampled category count K to 3 in adaptive Copy-Paste (ablation) | AEL with adaptive Copy-Paste · Cityscapes val 1/32 and 1/16 | mIoU: K=1: 72.18, 74.85; K=3: 74.28, 75.83 | Improved |
| Set unsupervised loss weight alpha to 1.0 (ablation) | AEL · Cityscapes val 1/32 and 1/16 | mIoU: alpha=0.5: 71.85, 74.61; alpha=1.0: 74.28, 75.83 | Improved |
Closest papers by meaning
- A Simple Baseline for Semi-Supervised Semantic Segmentation With Strong Data Augmentation (ICCV 2021 · similarity 0.90)
- Re-Distributing Biased Pseudo Labels for Semi-Supervised Semantic Segmentation: A Baseline Investigation (ICCV 2021 · similarity 0.90)
- Semi-supervised Semantic Segmentation with Prototype-based Consistency Regularization (NeurIPS 2022 · similarity 0.90)
- Semi-Supervised Semantic Segmentation With High- and Low-Level Consistency (TPAMI 2019 · similarity 0.89)
- Semi-Supervised Semantic Segmentation With Cross-Consistency Training (CVPR 2020 · similarity 0.88)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 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.
