V-Agent is a novel interactive video search system that integrates vision-language models (VLMs) to allow users to search video content using natural language queries. The system architecture combines a VLM for semantic understanding with a retrieval module for efficient indexing. In evaluations, V-Agent outperformed traditional keyword-based methods, achieving higher precision and recall. The interactive component enables iterative refinement of queries, enhancing user experience. This approach has implications for video surveillance, media archives, and content moderation. The paper, presented at ESWA 2025, underscores the growing role of VLMs in multimedia retrieval tasks. Developers and researchers can leverage this framework to build more intuitive video search tools.
V-Agent introduces an interactive video search system that leverages vision-language models to enable natural language queries. It demonstrates improved retrieval accuracy and user interaction. This work is significant for advancing video understanding and search technologies.