Video section importance calculation model learning device and its program, and summary video generator and its program

摘要

PROBLEM TO BE SOLVED: To provide a summary video generation device capable of generating a summary video without requiring special data other than video. SOLUTION: A summary video generation device 2 includes a video dividing means 210 for dividing an input video, a feature vector generating means 211 for generating an element-specific feature vector for each predetermined element for each divided section video, and a feature vector generating means 211. NN calculation means 212 that calculates the importance of each section video using a pre-learned neural network model that inputs the feature vector for each element and outputs the importance as a summary video, and the NN calculation means 212 in order from the one with the highest importance. It is provided with a section selection means 220 for selecting a section of a video within a predetermined time length, and a selection section connecting means 221 for generating a summary video by extracting the video of the selected section from the input video and concatenating the video. [Selection diagram] FIG. 6