EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Multi-level visual-language models feature learning for generalizable anomaly detection

jianfeng Qiu, junfa Li, Juan Xie, Xueliang Ma, Ke Xu

Complex & Intelligent Systems · 2027-01 · 13 (1)

DOI: 10.1007/s40747-026-02462-6 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Complex & Intelligent Systems, online erschienen am 17.08.2026. Verfasst von jianfeng Qiu und 4 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Zero-shot anomaly detection (ZSAD) aims to identify anomalies in target datasets without accessing their samples. Although CLIP and other large-scale vision language models show strong generalization, their potential for multi-level feature extraction in ZSAD remains underexplored. To address this, we propose a Multi-Level Feature Learning (MLFL) framework to enhance the zero-shot capability of CLIP via hierarchical alignment. MLFL adopts a two-stage training paradigm: Multi-Level Text Prompt Tuning (MLTP) and Multi-Level Text-Image Feature Alignment (MLFA). MLTP learns object-agnostic and object-aware prompts tailored to different encoder blocks. MLFA aligns textual and visual features using linear layers for shallow blocks and a Deep Feature Alignment (DFA) module for deep blocks. To compress parameters and preserve semantics, we introduce a Generalized Prompt Distillation (GPD) module that distills object-aware prompts into a unified representation. Experiments on seven industrial datasets achieve state-of-the-art performance, and deployment tests on edge devices demonstrate the potential applicability of the framework in practical industrial scenarios.

Förderung

  • National Natural Science Foundation

Zitieren

APA

Qiu, J., Li, J., Xie, J., Ma, X., & Xu, K. (2027). Multi-level visual-language models feature learning for generalizable anomaly detection. Complex & Intelligent Systems, 13(1). https://doi.org/10.1007/s40747-026-02462-6

Vancouver

Qiu J, Li J, Xie J, Ma X, Xu K. Multi-level visual-language models feature learning for generalizable anomaly detection. Complex & Intelligent Systems. 2027;13(1). doi:10.1007/s40747-026-02462-6

BibTeX
@article{Qiu2027,
  author = {Qiu, jianfeng and Li, junfa and Xie, Juan and Ma, Xueliang and Xu, Ke},
  title = {Multi-level visual-language models feature learning for generalizable anomaly detection},
  journal = {Complex & Intelligent Systems},
  year = {2027},
  volume = {13},
  number = {1},
  doi = {10.1007/s40747-026-02462-6}
}
Kostenlos starten →