Research map / Label-efficient and robust learning / Semi-supervised learning
A Simple Baseline for Semi-Supervised Semantic Segmentation With Strong Data Augmentation
Jianlong Yuan, Yifan Liu, Chunhua Shen, Zhibin Wang, Hao Li
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 self-training with pseudo-labels | DeepLabv3Plus · Cityscapes val (1/8 labeled) | mIoU: 68.9 -> 71.2 | Improved |
| add strong data augmentation to self-training (standard BN) (ablation) | DeepLabv3Plus ST · Cityscapes val (1/8 labeled) | mIoU: 71.2 -> 71.3 | No clear effect |
| add DSBN on top of strong augmentation (ablation) | DeepLabv3Plus ST+SDA · Cityscapes val (1/8 labeled) | mIoU: 71.3 -> 72.3 | Improved |
| use DSBN instead of fixed BN (ablation) | ST+SDA · Cityscapes val (1/8 labeled) | mIoU: 71.6 -> 72.3 | Improved |
| use DSBN instead of trainable shared BN (ablation) | ST+SDA · Cityscapes val (1/8 labeled) | mIoU: 71.3 -> 72.3 | Improved |
| add self-correction loss (ablation) | DeepLabv3Plus ST+SDA+DSBN · Cityscapes val (1/8 labeled) | mIoU: 72.3 -> 72.8 | Improved |
| add self-correction loss at 1/2 labeled (ablation) | DeepLabv3Plus ST+SDA+DSBN · Cityscapes val (1/2 labeled) | mIoU: 78.7 -> 78.7 | No clear effect |
| add iterative training (ablation) | DeepLabv3Plus · Cityscapes val (1/8 labeled) | mIoU: 72.8 -> 74.1 | Improved |
| add iterative training at 1/2 labeled (ablation) | DeepLabv3Plus · Cityscapes val (1/2 labeled) | mIoU: 78.7 -> 78.7 | No clear effect |
| add strong augmentation to fully labeled supervised training (ablation) | DeepLabv3Plus · Cityscapes val (full ground truth) | mIoU: 78.7 -> 78.7 | No clear effect |
| add strong augmentation to pseudo-label training (ablation) | DeepLabv3Plus · Cityscapes val (7/8 pseudo labels) | mIoU: 70.6 -> 71.4 | Improved |
| add DSBN to strong augmentation on pseudo labels (ablation) | DeepLabv3Plus · Cityscapes val (7/8 pseudo labels) | mIoU: 71.4 -> 72.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PseudoSeg — PseudoSeg: Designing Pseudo Labels for Semantic Segmentation (ICLR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Augmentation Matters: A Simple-Yet-Effective Approach to Semi-Supervised Semantic Segmentation (CVPR 2023 · similarity 0.93)
- Semi-Supervised Semantic Segmentation With High- and Low-Level Consistency (TPAMI 2019 · similarity 0.92)
- Semi-supervised Semantic Segmentation via Strong-weak Dual-branch Network (ECCV 2020 · similarity 0.91)
- Semi-supervised Semantic Segmentation with Prototype-based Consistency Regularization (NeurIPS 2022 · similarity 0.91)
- Re-Distributing Biased Pseudo Labels for Semi-Supervised Semantic Segmentation: A Baseline Investigation (ICCV 2021 · similarity 0.91)
- Universal Semi-Supervised Semantic Segmentation (ICCV 2019 · similarity 0.91)
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.
