Research map / Detection and segmentation / Weakly supervised detection and segmentation
Semi-supervised instance segmentation
Segment instances with mixed point, box, and unlabeled data via self-training.
16 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2020–2025) · an approach in Weakly supervised detection and segmentation, keyword cluster “semi supervised · pseudo · unlabeled”
State History Tried Checklist Preprints Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
8 papers in 2023–24 and 2 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 7.0% to 2.8%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2020 · Learning Saliency Propagation for Semi-Supervised Instance Segmentation (CVPR 2020 · 33 citations)
- 2021 · DatasetGAN: Efficient Labeled Data Factory With Minimal Human Effort (CVPR 2021 · 238 citations)
- 2022 · BigDatasetGAN: Synthesizing ImageNet With Pixel-Wise Annotations (CVPR 2022 · 75 citations)
- 2023 · The Devil Is in the Points: Weakly Semi-Supervised Instance Segmentation via Point-Guided Mask Representation (CVPR 2023 · 25 citations)
- 2024 · BSNet: Box-Supervised Simulation-assisted Mean Teacher for 3D Instance Segmentation (CVPR 2024 · 9 citations)
- 2025 · Iterative Self-Training with Class-Aware Text-to-Image Synthesis for Visual Task Learning (AAAI 2025 · 2 citations)
Earliest papers here
- Learning Saliency Propagation for Semi-Supervised Instance Segmentation (CVPR 2020 · 33 citations)
- DatasetGAN: Efficient Labeled Data Factory With Minimal Human Effort (CVPR 2021 · 238 citations)
- How to Save your Annotation Cost for Panoptic Segmentation? (AAAI 2021 · 2 citations)
What has been tried
16 papers here report 200 measured changes (a component added, removed or swapped, compared with a baseline): helped 80%, no clear effect 4%, hurt 14%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add local-structure attention | ScanNetV2 training overlapping areas · mAcc | 41.5 -> 48.1 | Helped | BSNet: Box-Supervised Simulation-assisted Mean Teacher for 3D Instance Segmentation (CVPR 2024 · 9 citations) |
| add initial extreme-point and background point supervision | PASCAL VOC val2012 · AP | 37.6 -> 38.0 | Helped | Extreme Point Supervised Instance Segmentation (CVPR 2024 · 8 citations) |
| add auxiliary semantic branch supervision (λsem>0) | VOC val, P1 · PQ | 32.9 -> 52.0 | Helped | Fully Data-Driven Pseudo Label Estimation for Pointly-Supervised Panoptic Segmentation (AAAI 2024 · 3 citations) |
| Use Chain Generation rather than Direct Generation prompts | PASCAL VOC 2012 · mIoU | 39.4 -> 48.2 | Helped | Iterative Self-Training with Class-Aware Text-to-Image Synthesis for Visual Task Learning (AAAI 2025 · 2 citations) |
| Use CascadePSP rather than CRF for pseudo-mask refinement | SA-1B · AR1000 | 15.3 -> 23.9 | Helped | Segment Anything without Supervision (NeurIPS 2024 · 2 citations) |
| Add high-precision retrieval and train with retrieved pseudo-labeled images | YouTube-VIS 2019 (2%) · mAP | 58.5% to 60.3% | Helped | Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation (CVPR 2025 · 1 citation) |
What reviewers will expect
Datasets and baselines named in the 6 papers of this approach from 2024–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- PASCAL VOC 2012 · 2 of 6 papers
- COCO · 2 of 6 papers
- SA-1B · 2 of 6 papers
Recent preprints
Loading recent arXiv preprints…
Use it
Download BibTeX of 16 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Pseudo-labeling for semi-supervised detection (76 papers)
- Consistency regularization for segmentation (25 papers)
Elsewhere in the topic
- Weakly supervised semantic segmentation (116 papers)
- Weakly supervised detection (97 papers)
- Weakly supervised object localization (54 papers)
- CAM refinement and prototypes (52 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).