Method and device for detecting violation image in video, equipment and storage medium

Abstract

The invention discloses a method and device for detecting violation images in a video, equipment and a storage medium. The method comprises: acquiring a current to-be-detected image in a video, wherein the to-be-detected image comprises a human face; determining a face score value of the to-be-detected image according to the distance between the feature vector of the face and a preset figure feature vector; and determining whether the to-be-detected image is a violation image or not according to the face score value. Compared with an existing violation image detection scheme, the method is advantageous in that the timeliness of violation image detection is effectively improved, and the investment of human resources is reduced, so that the cost consumption of video auditing is reduced, thedetection error of manual detection is reduced, the risk of missing detection of some sensitive persons is reduced, the auditing precision of the violation image in the video stream is ensured, and the method has better practicability.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a method, device, equipment and storage medium for detecting illegal images in videos. Background technique [0002] With the marketization of network information, many application products that focus on Internet video streams such as short videos or live broadcasts have gradually entered the markets of various countries. In addition to constraints, different current customs and habits need to be complied with. If some specific characters appear in the video uploaded by the application product at an inappropriate time or in an inappropriate way, it may cause local customs or even laws and regulations, thereby causing The discomfort of local users will affect the normal operation of application products in the local area. [0003] Generally speaking, a video containing a specific person uploaded in an application product can be regarded as an illegal video, while an ima...

AI Technical Summary

Problems solved by technology

[0004] At present, the common audit inspection is video audit based on artificial platform, that is, manual audit and judgment on video streams. Since Internet video streams have the characteristics of large data volume and high real-time requirements, this method has high cost investment, long audit time, and It is impossible to eliminate the shortcomings of error detection caused by human understanding deviation and artificial fatigue.

In addition, face recognition technology can also be used to identify and match specific people in the video stream to achieve illegal video inspection, but this method is limited by the face recognition accuracy of a single picture, which may cause some sensitive specific people to be missed. risks of

Images