매일 아침, 어제의 AI를 한 통으로 정리해 보내드립니다메일로 받아보기

METAL LAB

Glance, Scrutinize, and Think: Advancing Video Anomaly Detection from Training-Free to Agentic Reasoning

arXiv:2608.112602026-08-13

arXiv:2608.11260v1 Announce Type: new Abstract: Video Anomaly Detection (VAD) aims to identify anomalous events and localize their temporal intervals. Existing approaches exhibit a "when-what" dissociation: traditional DNN-based methods localize when anomalies occur but lack semantic understanding, whereas LLM-based methods explain what happens but neglect precise temporal grounding. We attribute this to the absence of a unified reasoning paradigm. Inspired by how humans inspect surveillance vid

저자 · Shibo Gao, Peipei Yang, Xu-Yao Zhang, Linlin Huang

arXiv에서 원문 보기