Video quality evaluation method and device, video processing device and medium

Abstract

The invention discloses a video quality evaluation method and device, a video processing device and a medium. The video quality evaluation method comprises the following steps: constructing a trainingdata set of an evaluation model to be used for a neural network system, wherein the training data set comprises manual annotation data and user click rate data; an attribute feature obtaining step: for each sample in the training data set, obtaining each attribute feature, related to the video attribute data, of each sample through a neural network system; a feature synthesis step: for each sample, performing feature synthesis processing on each attribute feature of the sample to obtain a synthesis feature of the sample; and a training step: performing confrontation training on the syntheticfeatures of each sample by combining the manual annotation task and the user click rate task, and generating an evaluation model for evaluating the video quality. According to the video quality evaluation model provided by the invention, a more accurate and effective evaluation result can be obtained for a new video or an exposed video.

Description

technical field [0001] The invention relates to video processing technology, in particular to a video quality evaluation method, equipment, video processing equipment and media. Background technique [0002] In the current era of mobile Internet, each user is not only a consumer of content, but also a creator of content, and various content generated by users has exploded. [0003] The inventor found that in the video recommendation scenario, only a small part of the massive videos generated by users can be displayed, and because the new videos appear for a short time, the accumulated clicks are often lower than the old videos, so the new videos are based on Clicks are also less likely to be recommended. [0004] In addition, the quality of content uploaded by users is also uneven. Low-quality videos are mainly reflected in unclear cover images and meaningless video descriptions. viewing experience. Therefore, in the video recommendation business, the quality of each vide...

AI Technical Summary

Problems solved by technology

Specifically, in the first method, since the marked data is manually judged, the judgment of whether the video is good or bad is highly subjective, and such data itself contains noise; in addition, since users themselves have personalized Therefore, it is difficult for high-quality videos marked by individual people to meet the personalized needs of a large number of users, that is, high-quality videos marked by a small number of people are not necessarily the ones that most users tend to click or watch.

In the second method, the positive and negative samples are extracted entirely based on the user's click volume, and because the old video has been online for a long time, the accumulated click volume is usually higher than that of the new video, so the user's click volume cannot be comprehensive. and accurately reflect the popularity of new and old videos, and not necessarily high-quality videos that users like to click

[0007] As mentioned above, for the quality evaluation of short videos, the quality score of the neural network model learned through the prior data of manual annotation or the posterior data of user behavior has a certain deviation.

Images