↓ Skip to main content
sis-arxiv-vad-papers

Chen Sun

2025

Personalizing Vision-Language Models With Hybrid Prompts for Zero-Shot Anomaly Detection

·8885 words·42 mins
Yunkang Cao ,  Xiaohao Xu ,  Yuqi Cheng ,  Chen Sun ,  Zongwei Du ,  Liang Gao ,  Weiming Shen
Cuhk-Avenue Shanghaitech Xd-Violence Ubnormal Ucf-Crime Ucsd-Ped Other Weakly Supervised Semi Supervised Training Free Instruction Tuning Unsupervised Hybrid Other
Introduces AnomalyVLM, a framework leveraging hybrid prompts derived from prior knowledge to enhance zero-shot anomaly detection by personalizing vision-language models, incorporating an anomaly region generator and refiner, and utilizing hybrid prompts for category-specific customization and improved detection performance.

2023

Towards Generic Anomaly Detection and Understanding: Large-scale Visual-linguistic Model (GPT-4V) Takes the Lead

·12272 words·58 mins
Yunkang Cao ,  Xiaohao Xu ,  Chen Sun ,  Xiaonan Huang ,  Weiming Shen
Hybrid Survey
This study explores the use of GPT-4V, a large visual-linguistic model, for generic anomaly detection across multiple modalities and domains, demonstrating its ability to understand global and fine-grained semantics, reason automatically, and improve with prompts. It evaluates GPT-4V on diverse tasks including industrial, medical, logical, video, 3D, and time series anomaly detection, discussing its promising performance and future directions for enhancement, such as quantitative metrics, expanded benchmarks, multi-round interactions, human feedback, and real-time application.