Design And Implementation Of A Video Highlight Positioning System Based On Deep Learning

With the popularity of smartphones and mobile networks in recent years,various video platforms have been developing rapidly.As a consequence,a large number of video clips are being generated every day,which leads to a series of problems such as insufficient video supervision,imprecise video recommendation,and inefficient video editing.The video highlight positioning system can use deep learning algorithms to locate the start time and end time of relevant actions in video clips,and thus owns great research significance and broad application prospect.Based on the background mentioned above,the main task is to engineer a system using deep learning algorithm for video highlight positioning.Specific points are as follows:1.The mainstream deep learning-based highlight positioning algorithms have been studied.The objectives and difficulties of each step have been analyzed.Moreover,various optional algorithm extensions were designed to assist users in quickly customizing the highlight positioning algorithm according to the task requirements and hardware constraints.2.A hybrid C/S and B/S structure based on deep learning for highlight positioning system were engineered.The system aims at supporting users to customize the algorithm process based on task requirements and hardware constraints through a graphical interactive interface.This system is also capable of parallel training,management of multiple tasks,dynamic loading of algorithm components,deployment of algorithm models and fast inference of input video clips.The experimental results show that the system can efficiently complete the design,training and reasoning of the highlight positioning algorithm through graphical interaction,and thus achieve the expected functional requirements.