Research map / Label-efficient and robust learning / Semi-supervised learning
Pseudo-SD: Pseudo Controlled Stable Diffusion for Semi-Supervised and Cross-Domain Semantic Segmentation
Dong Zhao, Qi Zang, Shuang Wang, Nicu Sebe, Zhun Zhong
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 reliable pseudo-labels only for partial attention manipulation instead of all pseudo-labels (ablation) | Pseudo-SD · Pascal (1/115) | mIoU: 74.9 -> 79.1 | Improved |
| Use reliable pseudo-labels only for partial attention manipulation instead of all pseudo-labels (ablation) | Pseudo-SD · COCO (1/512) | mIoU: 34.6 -> 39.9 | Improved |
| Use reliable pseudo-labels only for partial attention manipulation instead of all pseudo-labels (ablation) | Pseudo-SD · SFUDA G2C | mIoU: 58.7 -> 62.7 | Improved |
| Replace ControlNet with partial attention manipulation for fine-tuning SD (ablation) | Pseudo-SD · Pascal (1/115) | mIoU: 70.9 -> 79.1 | Improved |
| Replace ControlNet with partial attention manipulation for fine-tuning SD (ablation) | Pseudo-SD · SFUDA G2C | mIoU: 51.5 -> 62.7 | Improved |
| Replace ControlNet with partial attention manipulation for fine-tuning SD (ablation) | Pseudo-SD · UDA HRDA | mIoU: 65.8 -> 76.9 | Improved |
| Fine-tune SD with ControlNet on pseudo-labels instead of no synthetic data (ablation) | UniMatch / HRDA baselines · UDA HRDA | mIoU: 75.8 -> 50.4 | Hurt |
| Add Connected Components Removal (ablation) | Pseudo-SD · Pascal (1/115) | mIoU: 74.9 -> 79.1 | Improved |
| Add Connected Components Removal (ablation) | Pseudo-SD · SFUDA G2C | mIoU: 56.7 -> 62.7 | Improved |
| Add Hard-Adapted Class Resampling (ablation) | Pseudo-SD · Cityscapes (1/64) | mIoU: 67.1 -> 69.8 | Improved |
| Add Hard-Adapted Class Resampling (ablation) | Pseudo-SD · COCO (1/512) | mIoU: 37.1 -> 39.9 | Improved |
| Use unsupervised-segmentation structured pseudo-labels (STEGO) directly to train segmenter (ablation) | UniMatch · Pascal (1/115) | mIoU: 75.2 -> 73.3 | Hurt |
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.
- UniMatch — UniMatch: Universal Matching from Atom to Task for Few-Shot Drug Discovery (ICLR 2025) · better on 4 of 4 shared results
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
- Connectivity-Driven Pseudo-Labeling Makes Stronger Cross-Domain Segmenters (NeurIPS 2024 · similarity 0.90)
- Dataset Diffusion: Diffusion-based Synthetic Data Generation for Pixel-Level Semantic Segmentation (NeurIPS 2023 · similarity 0.90)
- PseudoSeg: Designing Pseudo Labels for Semantic Segmentation (ICLR 2021 · similarity 0.90)
- Re-Distributing Biased Pseudo Labels for Semi-Supervised Semantic Segmentation: A Baseline Investigation (ICCV 2021 · similarity 0.90)
- Unsupervised Modality Adaptation with Text-to-Image Diffusion Models for Semantic Segmentation (NeurIPS 2024 · similarity 0.88)
- Towards Robust Pseudo-Label Learning in Semantic Segmentation: An Encoding Perspective (NeurIPS 2025 · similarity 0.88)
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.
