This paper presents an information retrieval system for long-duration surveillance videos, utilizing an approach based on text processing generated by an object detection model. Our system offers a quick and efficient response, which can significantly improve the task of searching for scenes in surveillance videos. This tool has the potential to streamline security operations and facilitate the work of personnel responsible for monitoring and analyzing large volumes of visual data. However, more detailed evaluation and additional testing are required to validate its effectiveness in real surveillance environments. Ultimately, our goal is to contribute to the advancement of security and surveillance technologies through the development of innovative and practical solutions.

information retrival system, surveillance videos, objects identifiction, YOLO-NAS