Action Hints paper uses LLMs for skeleton-based video anomaly detection

Researchers have developed a new framework for zero-shot video anomaly detection (ZS-VAD) that leverages semantic typicality and context uniqueness from skeleton data. This approach aims to improve generalization to new scenes by distilling knowledge from large language models about normal and abnormal behaviors. The method achieves state-of-the-art results on multiple datasets without requiring target domain training data. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Enhances zero-shot video anomaly detection capabilities by leveraging LLM knowledge for skeleton-based analysis.

RANK_REASON This is a research paper introducing a novel framework for video anomaly detection.

Read on arXiv cs.CV →

• arXiv
• Canhui Tang
• ShanghaiTech
• UBnormal
• NWPU
• UCF-Crime

• paper
• other