Design and Implementation of Multimodal Video Retrieval System

With the application of deep learning in various fields, it replaces the extraction of video by manual design. Through the analysis and research of modal video retrieval methods, this paper aims to use text, image or video to carry out different ways of video retrieval through multi-modal video retrieval, and strive to meet the needs of different scenes, different users of video retrieval, and maximize the accuracy and effectiveness of video retrieval. This paper designs and implements a deep learning-based multimodal video retrieval system based on Windows. The system is based on the design concept of program modularization, using MySQL as the database and PyQT4 as the development tool of system interface. It mainly realizes three functional modules of video retrieval based on text, video retrieval based on image and video retrieval based on video segment.