Illegal video identification method and apparatus, and electronic device

Abstract

The embodiment of the invention provides a violation video recognition method and device and electronic equipment, and the method comprises the steps: recognizing a to-be-recognized video based on a preset recognition model, and obtaining a recognition result; wherein the identification model comprises a first identification model and a second identification model; wherein the identification result comprises a first identification result obtained based on the first identification model and a second identification result obtained based on the second identification model. If at least one of thefirst identification result and the second identification result is illegal, determining that the video is illegal; wherein the first identification model is obtained by training an initial convolutional neural network model by using initial sample data in advance, and the second identification model is obtained by training the first identification model by using update data of the initial sampledata in advance. Therefore, through the scheme, violation videos can be stably and reliably recognized without being influenced by long-time updating evolution, and the probability of missing detection is reduced.

Description

technical field [0001] The present invention relates to the technical field of video identification, in particular to an identification method, device and electronic equipment for illegal video. Background technique [0002] With the rapid development of the video industry, the number of videos faced by video supervision has exploded, and the identification method of manually watching each video to identify illegal videos has been difficult to meet the demand. At the same time, in view of the fact that video recognition is essentially a process of image recognition, computer vision technology is introduced to recognize video frames to realize automatic video recognition to meet the supervision needs of a large number of videos. [0003] In the video recognition technology, if the video is recognized frame by frame, a large amount of calculation is required, and the video recognition efficiency is too low. Therefore, most of the corresponding video recognition methods are ba...

AI Technical Summary

Problems solved by technology

[0004] However, the above-mentioned pre-trained convolutional neural network is often unable to identify some new violation videos that have not appeared in the sample data; and if new violation videos are included to update the original sample data, use the updated sample data To update the original convolutional neural network, it may cause the problem that after a long period of update and evolution, the convolutional neural network trained with updated sample data cannot identify the original illegal video well.

For example, in a certain period of time, the original violating videos do not appear for a long time, but a large number of new violating videos appear, so that the updated sample data contains far more new violating videos than the original violating videos, then use the updated sample data The trained convolutional neural network will not be able to recognize the original offending video very well

Images