Method and system for detecting deep forged video based on optical flow method

Abstract

The invention discloses a method and a system for detecting a deep forged video based on an optical flow method. The method comprises the following steps: collecting a deep forged video and an original video which are tampered in different modes to respectively construct a training set and a verification set; performing frame extraction processing on the videos in the training set and the verification set to obtain each frame of image of each video, and extracting optical flow features by using an optical flow calculation model; splicing each frame of image and the optical flow features according to a time sequence, and inputting a neural network model for training until loss convergence; carrying out frame extraction processing, optical flow calculation and image splicing on the video to be verified, taking part of the image and sending the image to a neural network model, obtaining an output result mean value through an activation function of a neural network according to obtained output, and if the output result mean value is not smaller than a preset threshold value, determining the video to be verified to be a deep forged video. According to the method, the strong dependence of a deep counterfeit detection algorithm on a preposition algorithm can be reduced, meanwhile, various deep counterfeit videos can be effectively detected, and the generalization ability of the algorithm is improved.

Description

technical field [0001] The present invention relates to the technical field of fake video detection, in particular to a method and system for deep fake video detection based on the optical flow method. Background technique [0002] Deep forgery technology refers to the use of deep learning related technologies to tamper with or generate video, audio, images, etc. Different from traditional forgery technologies, deepfakes are easy to generalize and difficult to distinguish with the naked eye, and the threshold for ordinary people to forge after obtaining a deepfake model is low. Therefore, it has developed rapidly in the past two years, and the identification methods for deepfake videos have also become increasingly popular. Came into being. Since the current deepfake videos are often spliced ​​by multiple frames of forged images, in view of this feature, the mainstream deepfake video identification methods can be divided into two types, one is based on artificial visual eff...

AI Technical Summary

Problems solved by technology

[0006] In order to solve the technical problems in the prior art that the detection method based on the inconsistency of biological characteristics fails, the inter-frame detection algorithm has a strong dependence on the pre-algorithm, and affects the generalization ability of the deep forgery detection algorithm, the present invention proposes a method based on optical flow Deep fake video detection method and system to solve the above technical problems

Images