Video summarizing method and device based on deep learning, and terminal equipment

Abstract

The invention discloses a video summarizing method based on intensive learning and deep learning, and the method comprises the steps: segmenting a target video, and obtaining a plurality of video segments; extracting a feature vector of all video frames of each video segment; taking the feature vector of the video frames of each video segments as the input, inputting the feature vectors into a pre-trained depth neural network, and obtaining the probability value corresponding to each outputted frame; calculating the importance value of each video segment according to the probability values corresponding to all video frames of the video segment; selecting a plurality of video segments with the bigger importance values from the video segments while guaranteeing that the proportion of the total duration of the selected video segments to the total duration of a target video is less than or equal to a preset proportion threshold value; carrying out the arrangement of the selected video segments, obtaining a video summary; defining an award function which represents the representativeness and the diversity of the video summary, and achieving the depth network training based on the non-supervision and supervision data through the intensive learning method.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a deep learning-based video summarization method, device and terminal equipment. Background technique [0002] In today's Internet-developed era, online videos on video websites emerge in endlessly. At the same time, videos often have a relatively long time span, which creates certain obstacles for users to browse quickly. This is because users cannot fully browse a large number of networks in a limited time. video. In order to facilitate users to understand the general content of the video in a short period of time before watching the complete video, it is an important research topic for those skilled in the art to find a method for quickly summarizing the video. Contents of the invention [0003] Embodiments of the present invention provide a deep learning-based video summarization method, device, and terminal device, which can quickly summarize videos and grea...

AI Technical Summary

Problems solved by technology

[0002] In today's Internet-developed era, online videos on video websites emerge in endlessly. At the same time, videos often have a relatively long time span, which creates certain obstacles for users to browse quickly. This is because users cannot fully browse a large number of networks in a limited time. video

Images