Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2601.22763 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918318556839936 |
|---|---|
| author | Zhang, Xingwu Li, Guanxuan Henderson, Paul Aragon-Camarasa, Gerardo Long, Zijun |
| author_facet | Zhang, Xingwu Li, Guanxuan Henderson, Paul Aragon-Camarasa, Gerardo Long, Zijun |
| contents | Current state-of-the-art multi-class unsupervised anomaly detection (MUAD) methods rely on training encoder-decoder models to reconstruct anomaly-free features. We first show these approaches have an inherent fidelity-stability dilemma in how they detect anomalies via reconstruction residuals. We then abandon the reconstruction paradigm entirely and propose Retrieval-based Anomaly Detection (RAD). RAD is a training-free approach that stores anomaly-free features in a memory and detects anomalies through multi-level retrieval, matching test patches against the memory. Experiments demonstrate that RAD achieves state-of-the-art performance across four established benchmarks (MVTec-AD, VisA, Real-IAD, 3D-ADAM) under both standard and few-shot settings. On MVTec-AD, RAD reaches 96.7\% Pixel AUROC with just a single anomaly-free image compared to 98.5\% of RAD's full-data performance. We further prove that retrieval-based scores theoretically upper-bound reconstruction-residual scores. Collectively, these findings overturn the assumption that MUAD requires task-specific training, showing that state-of-the-art anomaly detection is feasible with memory-based retrieval. Our code is available at https://github.com/longkukuhi/RAD. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_22763 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Is Training Necessary for Anomaly Detection? Zhang, Xingwu Li, Guanxuan Henderson, Paul Aragon-Camarasa, Gerardo Long, Zijun Computer Vision and Pattern Recognition Current state-of-the-art multi-class unsupervised anomaly detection (MUAD) methods rely on training encoder-decoder models to reconstruct anomaly-free features. We first show these approaches have an inherent fidelity-stability dilemma in how they detect anomalies via reconstruction residuals. We then abandon the reconstruction paradigm entirely and propose Retrieval-based Anomaly Detection (RAD). RAD is a training-free approach that stores anomaly-free features in a memory and detects anomalies through multi-level retrieval, matching test patches against the memory. Experiments demonstrate that RAD achieves state-of-the-art performance across four established benchmarks (MVTec-AD, VisA, Real-IAD, 3D-ADAM) under both standard and few-shot settings. On MVTec-AD, RAD reaches 96.7\% Pixel AUROC with just a single anomaly-free image compared to 98.5\% of RAD's full-data performance. We further prove that retrieval-based scores theoretically upper-bound reconstruction-residual scores. Collectively, these findings overturn the assumption that MUAD requires task-specific training, showing that state-of-the-art anomaly detection is feasible with memory-based retrieval. Our code is available at https://github.com/longkukuhi/RAD. |
| title | Is Training Necessary for Anomaly Detection? |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.22763 |