DINO-AD: Unsupervised Anomaly Detection with Frozen DINO-V3 Features
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915772551397376 |
|---|---|
| author | Huo, Jiayu Hong, Jingyuan Chen, Liyun |
| author_facet | Huo, Jiayu Hong, Jingyuan Chen, Liyun |
| contents | Unsupervised anomaly detection (AD) in medical images aims to identify abnormal regions without relying on pixel-level annotations, which is crucial for scalable and label-efficient diagnostic systems. In this paper, we propose a novel anomaly detection framework based on DINO-V3 representations, termed DINO-AD, which leverages self-supervised visual features for precise and interpretable anomaly localization. Specifically, we introduce an embedding similarity matching strategy to select a semantically aligned support image and a foreground-aware K-means clustering module to model the distribution of normal features. Anomaly maps are then computed by comparing the query features with clustered normal embeddings through cosine similarity. Experimental results on both the Brain and Liver datasets demonstrate that our method achieves superior quantitative performance compared with state-of-the-art approaches, achieving AUROC scores of up to 98.71. Qualitative results further confirm that our framework produces clearer and more accurate anomaly localization. Extensive ablation studies validate the effectiveness of each proposed component, highlighting the robustness and generalizability of our approach. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_03870 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | DINO-AD: Unsupervised Anomaly Detection with Frozen DINO-V3 Features Huo, Jiayu Hong, Jingyuan Chen, Liyun Image and Video Processing Computer Vision and Pattern Recognition Unsupervised anomaly detection (AD) in medical images aims to identify abnormal regions without relying on pixel-level annotations, which is crucial for scalable and label-efficient diagnostic systems. In this paper, we propose a novel anomaly detection framework based on DINO-V3 representations, termed DINO-AD, which leverages self-supervised visual features for precise and interpretable anomaly localization. Specifically, we introduce an embedding similarity matching strategy to select a semantically aligned support image and a foreground-aware K-means clustering module to model the distribution of normal features. Anomaly maps are then computed by comparing the query features with clustered normal embeddings through cosine similarity. Experimental results on both the Brain and Liver datasets demonstrate that our method achieves superior quantitative performance compared with state-of-the-art approaches, achieving AUROC scores of up to 98.71. Qualitative results further confirm that our framework produces clearer and more accurate anomaly localization. Extensive ablation studies validate the effectiveness of each proposed component, highlighting the robustness and generalizability of our approach. |
| title | DINO-AD: Unsupervised Anomaly Detection with Frozen DINO-V3 Features |
| topic | Image and Video Processing Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.03870 |