Research map / Detection and segmentation / Open-set and few-shot detection
Towards Open-World Object-based Anomaly Detection via Self-Supervised Outlier Synthesis
Brian Kostadinov Shalon Isaac-Medina, Yona Falinie Abdul Gaus, Neelanjan Bhowmik, Toby P Breckon
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 OLN-SSOS/FFS (pseudo-class outlier synthesis on class-agnostic detector) instead of class-based VOS/SIREN/FFS | object-level anomaly detection · DBF6 OoD test | OoD AR@100: 35.4 -> 48.8 | Improved |
| use OLN-SSOS/FFS instead of class-based baselines | object-level anomaly detection · SIXRay10 OoD test | OoD AR@100: 0.8 -> 35.6 | Improved |
| use OLN-FFS instead of best baseline SIREN | object-level anomaly detection · BDD100K / COCO OoD | OoD AR@100: 10.5 -> 15.9 | Improved |
| use OLN-FFS instead of VOS without class labels | object-level anomaly detection · PASCAL-VOC / COCO OoD | OoD AR@100: 20.6 -> 17.8 | Hurt |
| replace class-supervised detector with class-agnostic OLN and pseudo-classes | object-level anomaly detection · LTDImaging in-distribution | AR@100: 52.5 -> 17.8 | Hurt |
| add mask branch (OLN-Mask) to OLN-SSOS (ablation) | OLN-SSOS · DBF6 OoD test | OoD AR@100: 48.8 -> 58.9 | Improved |
| use FFS normalising-flow synthesis instead of Gaussian VOS synthesis (ablation) | OLN-SSOS · DBF6 and LTDImaging OoD | OoD AR@100: 58.9 -> 57.2 (DBF6 Mask); 18.2 -> 12.8 (LTD) | Mixed |
| increase number of pseudo-labels (ablation) | OLN-SSOS · DBF6, SIXRay10, VOC/COCO | AR@10/AR@100 | Improved |
| increase number of pseudo-labels (ablation) | OLN-SSOS · LTDImaging | AR@10/AR@100 | Hurt |
| outlier sampling size 300 versus other sizes (50 to 10,000) (ablation) | OLN-SSOS · VOC/COCO | AR@10/AR@100 | Improved |
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.
- PROB — PROB: Probabilistic Objectness for Open World Object Detection (CVPR 2023) · better on 1 of 1 shared result
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
- Open-World Objectness Modeling Unifies Novel Object Detection (CVPR 2025 · similarity 0.88)
- RUNA: Object-Level Out-of-Distribution Detection via Regional Uncertainty Alignment of Multimodal Representations (AAAI 2025 · similarity 0.88)
- Exploring Orthogonality in Open World Object Detection (CVPR 2024 · similarity 0.88)
- UMB: Understanding Model Behavior for Open-World Object Detection (NeurIPS 2024 · similarity 0.88)
- OW-OVD: Unified Open World and Open Vocabulary Object Detection (CVPR 2025 · similarity 0.87)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.87)
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.
