False face video identification method and system and readable storage medium

Abstract

The invention discloses a false face video identification method and system based on a convolutional neural network and an attention mechanism, and a readable storage medium. The method comprises thesteps: carrying out the sampling of an inputted video sequence, and obtaining N video frames; detecting, cutting and aligning the video frame to a human face to obtain a high-quality human face image,and obtaining a to-be-detected sample set; inputting the to-be-tested sample set into a trained deep convolutional neural network model of an attention mechanism to obtain an output result which is aresult of judging the authenticity of the video. According to the method disclosed in the invention, the negative sample is obtained through the special processing method of the positive sample, thetime cost of obtaining the negative sample is reduced, the face image of the fake face video is simulated well, and the trained network has good identification capability; in addition, the method canhighlight the manipulated image regions, thereby guiding the neural network to detect the regions, facilitating the detection of face counterfeiting, and improving the accuracy of an original CNN model.

Description

technical field [0001] The invention relates to the fields of computer vision and artificial intelligence, in particular to a method, system and readable storage medium for discriminating fake face videos based on a convolutional neural network and an attention mechanism. Background technique [0002] With the maturity of the application of deep learning technology, deepfake videos based on deepfake are becoming more and more popular. This deep-learning technique can create face-swapping videos that look like real ones. These fake videos look very realistic. So far, not only have these videos been faked in the context of pornography and verbal attacks to make people think that some celebrities are doing something discreditable to them, but even more egregiously, deepfake videos have also been used to impersonate political figures speech. U.S. Democratic Congressman Adam Schiff even warned that videos generated by Deepfake could have a disastrous impact on the 2020 U.S. el...

AI Technical Summary

Problems solved by technology

However, this method has its disadvantages. It uses artificial intelligence-based synthesis algorithms to generate fake images with low efficiency, requires high time and equipment costs, and the accuracy of identifying deep-face forged videos needs to be further improved.

Images