Research map / Label-efficient and robust learning / Semi-supervised learning
PseudoSeg: Designing Pseudo Labels for 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 PseudoSeg with unlabeled data instead of supervised-only training | DeepLabv3+ semantic segmentation · VOC12 low-data splits | mIoU | Helped |
| Use image-level labels in PseudoSeg instead of unlabeled data alone | PseudoSeg · VOC12 and COCO, all tested splits | mIoU | Helped |
| Add COCO training and unlabeled data | Supervised baseline · VOC12 val, full labeled training set | mIoU: 76.96 -> 77.40 (+0.44) | Helped |
| Add COCO training, unlabeled data, and VOC9k | Supervised baseline · VOC12 val, full labeled training set | mIoU: 76.96 -> 78.20 (+1.24) | Helped |
| Fuse decoder predictions with SGC maps using calibrated fusion (ablation) | PseudoSeg pseudo-label generation · VOC12 low-data splits, with and without image-level labels | mIoU | Helped |
| Replace calibrated fusion with decoder-only pseudo labels (ablation) | PseudoSeg pseudo-label generation · VOC12 low-data splits | mIoU | Hurt |
| Replace calibrated fusion with SGC-only pseudo labels (ablation) | PseudoSeg pseudo-label generation · VOC12 low-data splits | mIoU | Hurt |
| Use hypercolumn features instead of the last feature maps (ablation) | SGC pseudo-label refinement · VOC12 1/16 split | mIoU | Helped |
| Use all predictions as soft pseudo labels rather than selecting confident predictions (ablation) | PseudoSeg · VOC12 with image-level labels | mIoU | Helped |
| Apply temperature sharpening to pseudo-label distributions (ablation) | PseudoSeg · VOC12 with image-level labels | mIoU: 71.22 vs 68.15 | Helped |
| Use weaker color-jitter augmentation (ablation) | PseudoSeg strong augmentation · VOC12 with unlabeled data | mIoU | Hurt |
| Use geometric augmentations with STN alignment (ablation) | PseudoSeg consistency training · Semantic segmentation | mIoU | No clear effect |
Closest papers by meaning
- Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels (CVPR 2022 · similarity 0.92)
- Semi-Supervised Semantic Segmentation With Cross Pseudo Supervision (CVPR 2021 · similarity 0.91)
- Towards Robust Pseudo-Label Learning in Semantic Segmentation: An Encoding Perspective (NeurIPS 2025 · similarity 0.90)
- Improving Semi-Supervised Semantic Segmentation with Sliced-Wasserstein Feature Alignment and Uniformity (CVPR 2025 · similarity 0.89)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.89)
- RWMS: Reliable Weighted Multi-Phase for Semi-supervised Segmentation (AAAI 2024 · similarity 0.89)
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.