Research map / Label-efficient and robust learning / OOD and anomaly detection
Toward Long-Tailed Online Anomaly Detection through Class-Agnostic Concepts
Chiao-An Yang, Kuan-Chuan Peng, Raymond A. Yeh
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 |
|---|---|---|---|
| add S (semantic) branch to R branch on tail classes (ablation) | LTOAD · MVTec exp100 tail classes | image-level AUROC: 85.16 -> 88.14 | Improved |
| add prompt-conditioned generator G instead of feature jittering (ablation) | LTOAD · MVTec exp100 tail classes | image-level AUROC: 85.63 -> 88.14 | Improved |
| add the R branch (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 54.97 -> 93.72 | Improved |
| learn concepts in-domain instead of random out-domain concepts (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 91.54 -> 93.72 | Improved |
| use 10 learned concepts instead of 3 (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 91.83 -> 93.72 | Improved |
| learn concepts instead of fixing them (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 90.94 -> 93.72 | Improved |
| learn abnormal prompts instead of fixed (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 92.06 -> 93.72 | Improved |
| use multiple abnormal prompts per concept instead of single (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 91.92 -> 93.72 | Improved |
| use concept-specific abnormal prompts instead of shared (ablation) | LTOAD · MVTec exp100 (offline), all classes | image-level AUROC: 92.78 -> 93.72 | Improved |
| use anomaly adaptive online algorithm instead of naive online (ablation) | LTOAD · MVTec online (avg of 8 stream configs) | pixel-level AUROC: 95.08 -> 95.28 | Improved |
| go class-agnostic vs class-aware LTOAD | LTOAD · MVTec exp100 offline | image-level AUROC: 93.12 -> 93.42 | No clear effect |
| online learning with A^AA in cross-dataset MVTec to VisA | LTOAD · MVTec->VisA cross-dataset | pixel-level AUROC: 84.38 -> 92.37 | 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.
- MoEAD — MoEAD: A Parameter-efficient Model for Multi-class Anomaly Detection (ECCV 2024) · better on 3 of 3 shared results
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
- Long-Tailed Anomaly Detection with Learnable Class Names (CVPR 2024 · similarity 0.94)
- ADPretrain: Advancing Industrial Anomaly Detection via Anomaly Representation Pretraining (NeurIPS 2025 · similarity 0.88)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.88)
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.87)
- TailedCore: Few-Shot Sampling for Unsupervised Long-Tail Noisy Anomaly Detection (CVPR 2025 · similarity 0.87)
- Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · similarity 0.86)
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.
