Research map / Label-efficient and robust learning / Semi-supervised learning
Negative Pseudo Labeling using Class Proportion for Semantic Segmentation in Pathology
Hiroki Tokunaga, Brian Kenji Iwana, Yuki Teramoto, Akihiko Yoshizawa, Ryoma Bise
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 negative pseudo labels and adaptive positive threshold (full method) | Pseudo Labeling · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.420 -> 0.435 | Improved |
| use weak proportion labels with pseudo labeling vs supervised only | Supervised MobileNetV2 · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.373 -> 0.435 | Improved |
| use proposed method instead of ICT/Mean Teachers/mixup/VAT | SSL baselines · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.270 to 0.324 -> 0.435 | Improved |
| standard SSL methods using unlabeled data vs supervised only | ICT, Mean Teachers, VAT · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.373 -> 0.289 | Hurt |
| replace fixed 0.95 threshold with adaptive positive threshold (ablation) | Pseudo Labeling · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.420 -> 0.412 | Hurt |
| add single negative labeling to adaptive positive labeling (ablation) | Adaptive Pseudo Labeling · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.412 -> 0.426 | Improved |
| add multi negative labeling to adaptive positive labeling (ablation) | Adaptive Pseudo Labeling · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.412 -> 0.435 | Improved |
| use multi negative instead of single negative labeling (ablation) | Adaptive + negative labeling · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.426 -> 0.435 | Improved |
| add positive pseudo labels to single negative labeling (ablation) | Negative labeling only · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.348 -> 0.426 | Improved |
| add positive pseudo labels to multi negative labeling (ablation) | Negative labeling only · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.336 -> 0.435 | Improved |
| negative pseudo labeling alone vs supervised (ablation) | Single negative label only · Lung adenocarcinoma subtype segmentation (5-fold CV) | mIoU: 0.373 -> 0.348 | Hurt |
Closest papers by meaning
- Prototype-Based Image Prompting for Weakly Supervised Histopathological Image Segmentation (CVPR 2025 · similarity 0.86)
- Label-Efficient Hybrid-Supervised Learning for Medical Image Segmentation (AAAI 2022 · similarity 0.86)
- Weakly Supervised Gland Segmentation with Class Semantic Consistency and Purified Labels Filtration (AAAI 2025 · similarity 0.86)
- Minimizing Labeling Cost for Nuclei Instance Segmentation and Classification with Cross-domain Images and Weak Labels (AAAI 2021 · similarity 0.86)
- A Semantic Knowledge Complementarity based Decoupling Framework for Semi-supervised Class-imbalanced Medical Image Segmentation (CVPR 2025 · similarity 0.84)
- Annotation Ambiguity Aware Semi-Supervised Medical Image Segmentation (CVPR 2025 · similarity 0.84)
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.
