AA-CLIP: Enhancing Zero-shot Anomaly Detection via Anomaly-Aware CLIP
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866916647879573504 |
|---|---|
| author | Ma, Wenxin Zhang, Xu Yao, Qingsong Tang, Fenghe Wu, Chenxu Li, Yingtai Yan, Rui Jiang, Zihang Zhou, S. Kevin |
| author_facet | Ma, Wenxin Zhang, Xu Yao, Qingsong Tang, Fenghe Wu, Chenxu Li, Yingtai Yan, Rui Jiang, Zihang Zhou, S. Kevin |
| contents | Anomaly detection (AD) identifies outliers for applications like defect and lesion detection. While CLIP shows promise for zero-shot AD tasks due to its strong generalization capabilities, its inherent Anomaly-Unawareness leads to limited discrimination between normal and abnormal features. To address this problem, we propose Anomaly-Aware CLIP (AA-CLIP), which enhances CLIP's anomaly discrimination ability in both text and visual spaces while preserving its generalization capability. AA-CLIP is achieved through a straightforward yet effective two-stage approach: it first creates anomaly-aware text anchors to differentiate normal and abnormal semantics clearly, then aligns patch-level visual features with these anchors for precise anomaly localization. This two-stage strategy, with the help of residual adapters, gradually adapts CLIP in a controlled manner, achieving effective AD while maintaining CLIP's class knowledge. Extensive experiments validate AA-CLIP as a resource-efficient solution for zero-shot AD tasks, achieving state-of-the-art results in industrial and medical applications. The code is available at https://github.com/Mwxinnn/AA-CLIP. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_06661 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AA-CLIP: Enhancing Zero-shot Anomaly Detection via Anomaly-Aware CLIP Ma, Wenxin Zhang, Xu Yao, Qingsong Tang, Fenghe Wu, Chenxu Li, Yingtai Yan, Rui Jiang, Zihang Zhou, S. Kevin Computer Vision and Pattern Recognition Artificial Intelligence Anomaly detection (AD) identifies outliers for applications like defect and lesion detection. While CLIP shows promise for zero-shot AD tasks due to its strong generalization capabilities, its inherent Anomaly-Unawareness leads to limited discrimination between normal and abnormal features. To address this problem, we propose Anomaly-Aware CLIP (AA-CLIP), which enhances CLIP's anomaly discrimination ability in both text and visual spaces while preserving its generalization capability. AA-CLIP is achieved through a straightforward yet effective two-stage approach: it first creates anomaly-aware text anchors to differentiate normal and abnormal semantics clearly, then aligns patch-level visual features with these anchors for precise anomaly localization. This two-stage strategy, with the help of residual adapters, gradually adapts CLIP in a controlled manner, achieving effective AD while maintaining CLIP's class knowledge. Extensive experiments validate AA-CLIP as a resource-efficient solution for zero-shot AD tasks, achieving state-of-the-art results in industrial and medical applications. The code is available at https://github.com/Mwxinnn/AA-CLIP. |
| title | AA-CLIP: Enhancing Zero-shot Anomaly Detection via Anomaly-Aware CLIP |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2503.06661 |