Processing method for quickly auditing video content

Abstract

The invention provides a processing method for quickly auditing video content. The processing method comprises the following steps: uploading to-be-audited batch video data to a data storage module; the data processing strategy configuration module configures a data processing mode, configures a quantitative recognition threshold value to determine the number of video frames of one-time recognition, and configures a termination threshold value to determine to terminate a recognition task in advance; performing unified decoding, frame extraction and duplicate removal on the video; checking whether the video recognition result exists in the cache service or not, and if yes, directly returning the result; and if the video frame data does not exist, inputting the decoded and deduplicated data into an identification module, carrying out violation type identification on the video frame data in batches, sending an identification result to a cache service, and returning the result. According to the method, a batch identification and advanced ending strategy can be configured, various identification models are combined, content auditing of a large number of videos is automatically completed, and user usability is greatly facilitated; the recognition model greatly improves the recognition accuracy and speed, and greatly improves the auditing efficiency.

Description

technical field [0001] The invention relates to the technical field of Internet video processing, in particular to a processing method for increasing the review speed of video content. Background technique [0002] With the upsurge of the big data era, the development of Internet technology is booming, and there are more and more video resources on the Internet. This has resulted in a variety of videos on the Internet, and the content is mixed. For example, some video clips contain pornographic, violent, politically sensitive, and copyright-infringing content. It is particularly important to protect the Internet environment and protect the rights and interests of video copyright owners. Therefore, the content of the video needs to be reviewed. [0003] At present, the number of Internet users is increasing day by day, and the number of clicks, playbacks and reposts of a single video can reach tens of millions or even hundreds of millions of times. However, the current au...

AI Technical Summary

Problems solved by technology

However, the current review video technology, whether it is manual or system review, is mostly a review video from the beginning to the end of the system, including a large number of repeated videos, and video decoding review one by one

However, the general video length is tens of minutes or even one or two hours. Under the huge number of videos, the efficiency of this review method is insignificant.

[0004] The current deep learning recognition model is a single classification model, which can only rely on a single classification model to determine whether the video is a violation type. For some specific or complex content violation videos, the recognition accuracy is not high, resulting in the violation video miss treatment, invalid treatment

Significantly wasted processing time, further reducing efficiency

Images