Research map / Detection and segmentation / Open-set and few-shot detection
Object-centric learning theory and video
Learn object-centric representations with identifiability, robustness analysis, and video conditioning.
86 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in Open-set and few-shot detection, keyword cluster “object centric · slot · representations”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
20 papers in 2023–24 and 27 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 3.4% to 4.9%.
No longer reported by papers here: CLEVRTEX (mIoU, last in 2024); Multi-dSprites (ARI (%), last in 2023).
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
- 2018 · Adversarial Scene Editing: Automatic Object Removal from Weak Supervision (NeurIPS 2018 · 51 citations)
- 2019 · Object-Centric Auto-Encoders and Dummy Anomalies for Abnormal Event Detection in Video (CVPR 2019 · 432 citations)
- 2020 · Demystifying Contrastive Self-Supervised Learning: Invariances, Augmentations and Dataset Biases (NeurIPS 2020 · 128 citations)
- 2021 · PiCIE: Unsupervised Semantic Segmentation Using Invariance and Equivariance in Clustering (CVPR 2021 · 156 citations)
- 2022 · ObjectFolder 2.0: A Multisensory Object Dataset for Sim2Real Transfer (CVPR 2022 · 67 citations)
- 2023 · Bridging the Gap to Real-World Object-Centric Learning (ICLR 2023 · 30 citations)
- 2024 · Prompt-Driven Dynamic Object-Centric Learning for Single Domain Generalization (CVPR 2024 · 26 citations)
- 2025 · Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · 4 citations)
- 2026 · Causal-JEPA: Learning World Models through Object-Level Latent Masking (ICML 2026 · 5 citations)
Earliest papers here
- Adversarial Scene Editing: Automatic Object Removal from Weak Supervision (NeurIPS 2018 · 51 citations)
- Learning to Rank Based on Analogical Reasoning (AAAI 2018 · 15 citations)
- Object-Centric Auto-Encoders and Dummy Anomalies for Abnormal Event Detection in Video (CVPR 2019 · 432 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
MOVi-E · FG-ARI (higher is better)
5 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 84.1 | SAVi++: Towards End-to-End Object-Centric Learning from Real-World Videos (NeurIPS 2022 · 26 citations) |
CLEVRTEX · mIoU (higher is better)
4 papers, 2023–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 42.4 | Differentiable Mathematical Programming for Object-Centric Representation Learning (ICLR 2023 · 0 citations) |
| 2024 | 75.3 | Learning to Compose: Improving Object Centric Learning by Injecting Compositionality (ICLR 2024 · 0 citations) |
MOVi-C · FG-ARI (higher is better)
4 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 81.9 | SAVi++: Towards End-to-End Object-Centric Learning from Real-World Videos (NeurIPS 2022 · 26 citations) |
Multi-dSprites · ARI (%) (higher is better)
4 papers, 2021–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 91.2 | Efficient Iterative Amortized Inference for Learning Symmetric and Disentangled Multi-Object Representations (ICML 2021 · 3 citations) |
| 2022 | 95 | Evaluating Disentanglement of Structured Representations (ICLR 2022 · 1 citation) |
| 2023 | 99.8 | Multi-Object Representation Learning via Feature Connectivity and Object-Centric Regularization (NeurIPS 2023 · 0 citations) |
What has been tried
72 papers here report 789 measured changes (a component added, removed or swapped, compared with a baseline): helped 67%, no clear effect 9%, hurt 22%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add dynamic gating network (no prompt, no Slot-Attention) over static ResNet-18 base | PACS single-DG · Avg accuracy (%) | 54.97 -> 64.27 | Helped | Prompt-Driven Dynamic Object-Centric Learning for Single Domain Generalization (CVPR 2024 · 26 citations) |
| Add object-level contrastive learning to correspondence distillation | COCO-Stuff · unsupervised Acc. | 46.9 -> 64.2 | Helped | EAGLE: Eigen Aggregation Learning for Object-Centric Unsupervised Semantic Segmentation (CVPR 2024 · 20 citations) |
| Add instance-aligned semantics to vanilla semantics + SF2SE3 | Cityscapes val pseudo labels · PQ | 14.3 -> 14.9 | Helped | Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · 4 citations) |
| add the Adaptive Background Preservation loss (ABP) | MSCOCO 2017 ablation · L1B | 0.48 -> 0.10 | Helped | Style-Editor: Text-driven Object-centric Style Editing (CVPR 2025 · 2 citations) |
| Add object feature learning (OFL) to SGPN | 3DSSG · Object R@1 | 47.37 -> 54.49; +7.12% | Helped | Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · 2 citations) |
| Add masking | ADE20K · mIoU | 47.4 -> 48.6 | Helped | A Data-Centric Revisit of Pre-Trained Vision Models for Robot Learning (CVPR 2025 · 1 citation) |
| Remove object-centric representations | SAM segmentation benchmarks · mIoU | Hurt | OCRT: Boosting Foundation Models in the Open World with Object-Concept-Relation Triad (CVPR 2025 · 1 citation) | |
| add region combination after fusion | PASCAL VOC 2012 · mBOi | 45.4 -> 45.2 | Hurt | Pay Attention to the Foreground in Object-Centric Learning (CVPR 2025 · 1 citation) |
| Increase concept capacity by reducing SpLiCE sparsity regularization to match OCB | PASCAL-VOC, COCO(h), COCO(l), SUN397, COCOLogic · mAP / accuracy / balanced accuracy | 82.42 ± 0.01 -> 82.90 ± 0.01; 84.73 ± 0.00 -> 86.31 ± 0.00; 59.19 ± 0.00 -> 61.70 ± 0.00; 74.79 ± 0.01 -> 74.85 ± 0.01; | Helped | Object-Centric Concept-Bottlenecks (NeurIPS 2025 · 1 citation) |
| Use six slot-attention iterations instead of the default three | COCO · FG-ARI | 28.5 | Hurt | Bootstrapping Top-down Information for Self-modulating Slot Attention (NeurIPS 2024 · 1 citation) |
| add OLIP's full-image attention query | Object realism test set — Joint · Accuracy | 0.479 -> 0.539 | Helped | Be More Specific: Evaluating Object-centric Realism in Synthetic Images (CVPR 2025 · 0 citations) |
| replace compactness-based anchor selection with random anchor-node selection | ObjectsRoom, decoder foreground · ARI | 0.894 -> 0.832 | Hurt | Hierarchical Compact Clustering Attention (COCA) for Unsupervised Object-Centric Learning (CVPR 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 21 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ImageNet · 5 of 21 papers
- COCO · 5 of 21 papers
- VOC · 3 of 21 papers
- KITTI · 3 of 21 papers
- Open Images · 2 of 21 papers
- LVIS · 2 of 21 papers
- Tetrominoes · 2 of 21 papers
- CLEVR6 · 2 of 21 papers
Baselines
- DINOSAUR · 5 of 21 papers
- SlotDiffusion · 3 of 21 papers
- GENESIS-v2 · 2 of 21 papers
- BOQ-SA · 2 of 21 papers
- Slot Attention · 2 of 21 papers
- SLATE · 2 of 21 papers
- SPOT · 2 of 21 papers
- CutLER · 2 of 21 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 47 papers; from OpenAlex and the papers' affiliations).
- Max Planck Institute for Intelligent Systems · 3 papers
- Chinese Academy of Sciences · 2 papers
- University of Edinburgh · 2 papers
- Koc University · 2 papers
- Max-Planck Institute · 2 papers
- University of Tuebingen · 2 papers
- National University of Singapore · 2 papers
- Shanghai Jiao Tong University · 2 papers
Use it
Download BibTeX of 38 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
- Slot attention methods (20 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).