Video auditing method and device, auditing server and storage medium

Abstract

The invention discloses a video auditing method and device, an auditing server and a storage medium. The method comprises the following steps: inputting each key video frame in a to-be-audited video into a violation classification model, and obtaining a classification score of each key video frame under a corresponding violation sub-class through a binary classification module under different violation sub-classes in the violation classification model; for each violation sub-category, fusing the classification score of each key video frame under the violation sub-category to obtain a violationscore of the to-be-audited video under the violation sub-category; and determining violation category composition of the to-be-audited video according to the violation scores of the to-be-audited video under different violation sub-categories and a preset violation threshold. According to the technical scheme provided by the invention, multi-violation category judgment of the to-be-audited videois realized, the problem of misjudgment or missed judgment of violation sub-categories is avoided, the audit independence of the to-be-audited video under different violation sub-categories is ensured, and the comprehensiveness and accuracy of video audit are improved.

Description

technical field [0001] The embodiments of the present invention relate to the field of auditing illegal content in Internet resources, and in particular to a video auditing method, device, auditing server and storage medium. Background technique [0002] With the rapid development of Internet technology, streaming media resources on the Internet have grown explosively, making various short videos or webcast videos and other new user-generated content (User Generated Content, UGC) spread on the Internet more and more abundant. At the same time, a large number of illegal videos involving topics such as terrorism, violence, pornography, or political sensitivity are also produced and quickly spread on the Internet. Therefore, for various types of Internet videos, it is necessary to pre-judge the type of violation to prevent any violation Types of dissemination of Internet video. [0003] At present, video review methods usually use the Convolution Neural Network (CNN) model to ...

AI Technical Summary

Problems solved by technology

[0003] At present, the video review method usually uses the Convolution Neural Network (CNN) model to extract and train the features of each video frame in the Internet video, and uses the trained neural network model to sequentially determine whether there is a violation in a single video frame. Content, and then judge whether the entire Internet video violates the regulations based on the detection results of multiple single video frames. At this time, it is impossible to judge whether the video violates the regulations through the feature correlation relationship between multiple video frames, which can easily cause misjudgment or missed judgment. The review method of mutually exclusive classification can only judge whether the video is normal or violates the regulations, but cannot accurately review the specific violation type of the video when it violates the regulations

Images