Research map / Detection and segmentation / Open-set and few-shot detection
Zero-shot segmentation and detection
Recognize unseen classes via generative features and semantic alignment.
49 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Open-set and few-shot detection, keyword cluster “zero · shot object · shot segmentation”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
24 papers in 2023–24 and 8 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 4.1% to 1.4%.
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
- 2016 · Synthesized Classifiers for Zero-Shot Learning (CVPR 2016 · 713 citations)
- 2017 · DECK: Discovering Event Composition Knowledge from Web Images for Zero-Shot Event Detection and Recounting in Videos (AAAI 2017 · 16 citations)
- 2019 · Zero-Shot Object Detection with Textual Descriptions (AAAI 2019 · 80 citations)
- 2020 · Learning the Redundancy-Free Features for Generalized Zero-Shot Object Recognition (CVPR 2020 · 113 citations)
- 2021 · Goal-Oriented Gaze Estimation for Zero-Shot Learning (CVPR 2021 · 166 citations)
- 2022 · Disentangling Visual Embeddings for Attributes and Objects (CVPR 2022 · 82 citations)
- 2023 · ZegCLIP: Towards Adapting CLIP for Zero-Shot Semantic Segmentation (CVPR 2023 · 233 citations)
- 2024 · AnyDoor: Zero-shot Object-level Image Customization (CVPR 2024 · 175 citations)
- 2025 · Towards Efficient Foundation Model for Zero-shot Amodal Segmentation (CVPR 2025 · 6 citations)
- 2026 · OOVDet: Low-Density Prior Learning for Zero-Shot Out-of-Vocabulary Object Detection (ICML 2026 · 0 citations)
Earliest papers here
- Synthesized Classifiers for Zero-Shot Learning (CVPR 2016 · 713 citations)
- DECK: Discovering Event Composition Knowledge from Web Images for Zero-Shot Event Detection and Recounting in Videos (AAAI 2017 · 16 citations)
- Zero-Shot Object Detection with Textual Descriptions (AAAI 2019 · 80 citations)
What has been tried
44 papers here report 489 measured changes (a component added, removed or swapped, compared with a baseline): helped 77%, no clear effect 3%, hurt 16%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| use DINOv2 instead of VGG/CLIP as ID extractor backbone | DreamBooth benchmark · DINO Score | DINO 27.7 (VGG) / 31.5 (CLIP G+P) -> 64.1 (DINOv2 G+P) | Helped | AnyDoor: Zero-shot Object-level Image Customization (CVPR 2024 · 175 citations) |
| Aggregate attention maps proportionally across resolutions rather than using only 64×64 maps | segmentation examples · segmentation quality | Helped | Diffuse Attend and Segment: Unsupervised Zero-Shot Segmentation using Stable Diffusion (CVPR 2024 · 79 citations) | |
| add depth prediction supervision to the two-stream model | DAVIS-16, FBMS, and Long-Videos · J | 75.9 -> 77.0; 75.1 -> 77.5; 63.9 -> 62.8 | Mixed | Depth-aware Test-Time Training for Zero-shot Video Object Segmentation (CVPR 2024 · 41 citations) |
| use object-agnostic rather than object-aware text prompts | ZSAD image- and pixel-level datasets · AUROC | Helped | AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (ICLR 2024 · 32 citations) | |
| add re-perception | MP3D · SR | 36.5 -> 40.1 | Helped | SG-Nav: Online 3D Scene Graph Prompting for LLM-based Zero-shot Object Navigation (NeurIPS 2024 · 30 citations) |
| add the region-level bridge | PASCAL VOC 2012 · hIoU | 84.3 -> 85.4 | Helped | Exploring Regional Clues in CLIP for Zero-Shot Semantic Segmentation (CVPR 2024 · 25 citations) |
| Add primitive-wise attribute and object learning to composition-wise learning | MIT-States CW · AUC | 19.1 -> 21.1 | Helped | Context-based and Diversity-driven Specificity in Compositional Zero-Shot Learning (CVPR 2024 · 21 citations) |
| add object-aware feature grouping | Real-world colon tumor · DSC | 31.28 -> 36.23 | Helped | ZePT: Zero-Shot Pan-Tumor Segmentation via Query-Disentangling and Self-Prompting (CVPR 2024 · 20 citations) |
| make matching bidirectional rather than one-way from t to t+1 | LEVIR-CD · mask AR@1000 | 1.3 -> 32.6 | Helped | Segment Any Change (NeurIPS 2024 · 19 citations) |
| add SPT | PASCAL VOC 2012 · mIoU(U) | 77.8 -> 86.7 | Helped | Spectral Prompt Tuning: Unveiling Unseen Classes for Zero-Shot Semantic Segmentation (AAAI 2024 · 16 citations) |
| Add class-agnostic training | COCO-stuff · mIoU | 45.99 -> 49.17 | Helped | CSL: Class-Agnostic Structure-Constrained Learning for Segmentation Including the Unseen (AAAI 2024 · 12 citations) |
| add the separation-to-fusion structure | COCOA-cls · mean IoU (%) | 78.08 -> 80.54 | Helped | Towards Efficient Foundation Model for Zero-shot Amodal Segmentation (CVPR 2025 · 6 citations) |
What reviewers will expect
Datasets and baselines named in the 18 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 · 3 of 18 papers
- Cityscapes · 2 of 18 papers
- COCO-Stuff 164K · 2 of 18 papers
- YouTubeVOS · 2 of 18 papers
- MIT-States · 2 of 18 papers
- C-GQA · 2 of 18 papers
- HM3D · 2 of 18 papers
Baselines
- ZegFormer · 4 of 18 papers
- ZSSeg · 4 of 18 papers
- ZegCLIP · 3 of 18 papers
- CLIP · 3 of 18 papers
- SML · 2 of 18 papers
- SPNet · 2 of 18 papers
- ZS3 · 2 of 18 papers
- CaGNet · 2 of 18 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 32 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 3 papers
- Zhejiang University · 3 papers
- Google · 2 papers
- Beijing University of Posts and Telecommunications · 2 papers
- The University of Hong Kong · 2 papers
- Nanyang Technological University · 2 papers
Use it
Download BibTeX of 36 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
- Few-shot object detection (67 papers)
- Few-shot segmentation (64 papers)
- Prompted SAM segmentation (39 papers)
Elsewhere in the topic
- Object-level representation learning (219 papers)
- Interactive and referring segmentation (161 papers)
- Open-set and unknown object detection (160 papers)
- Amodal and open-world instance segmentation (137 papers)
- Industrial anomaly detection (117 papers)
- Unsupervised object and part discovery (114 papers)
- Domain-adaptive object detection (103 papers)
- Open-vocabulary object detection (96 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).