New framework bridges pose-semantic gap for text-based person anomaly search

Researchers have developed a new framework called the Structure-Semantic Decoupled Cascade (SSDC) to improve text-based person anomaly search in surveillance footage. This method addresses the Pose-Semantic Gap, where different actions can have similar skeletal structures, by first using a lightweight model for coarse retrieval based on skeletal similarity. It then employs a multi-agent semantic verification module, comprising a Detective, Analyst, and Writer, to refine the search results by synthesizing semantic captions and fusing them with structural information. AI

IMPACT Introduces a novel framework for efficient and accurate anomaly detection in surveillance, potentially improving security systems.

RANK_REASON This is a research paper introducing a new framework for anomaly search in computer vision.

Read on arXiv cs.CV →

• arXiv
• PAB benchmark

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →