Open access
Aug 2026
Multi-level visual-language models feature learning for generalizable anomaly detection
A Multi-Level Feature Learning (MLFL) framework is proposed to enhance the zero-shot capability of CLIP via hierarchical alignment and introduces a Generalized Prompt Distillation (GPD) module that distills object-aware prompts into a unified representation.
Jianfeng Qiu, Junfa Li, Juan Xie et al.
· Complex & Intelligent Sy... · 0 citations