Synthetic face image evidence obtaining method based on local binary pattern and deep learning

Abstract

The invention discloses a synthetic face image evidence obtaining method based on a local binary pattern and deep learning, and the method comprises the steps: collecting and marking a real face imageand a synthetic face image, creating a face image evidence obtaining sample image library, and dividing the image library into three parts: a training set, a verification set and a test set; determining an LBP operator mode and a sampling radius according to the extracted local binary pattern LBP features of the face image; constructing a face forensic convolutional neural network model, settinga convolutional neural network training hyper-parameter, obtaining an evaluation score according to logistic regression in a classification module, and updating a network parameter in the feature extraction module according to a loss function, a data label and the evaluation score; and training a neural network through the training set and the test set to obtain a training model, wherein the training model detects whether the input face image is a real natural face or a synthetic face. According to the invention, the common synthetic face image at the present stage can be rapidly and efficiently detected.

Description

Technical field [0001] The invention belongs to the technical field of machine learning and image forensics, and specifically relates to a synthetic face image forensics method based on local binary mode and deep learning. Background technique [0002] In recent years, the rapid development of computer vision technology and deep learning technology has made the editing and synthesis of face images easier and easier. While enriching people’s entertainment life, the masses of false synthetic faces in the media have also brought the public to life. Here comes a crisis of confidence. Once false face images are used maliciously, such as making fake news to mislead the public, using synthetic faces for identification, or as perjury in court, distorting facts, etc., it will lead to serious consequences. However, the advancement of synthesis technology makes the synthesized images more and more realistic, and people can no longer accurately judge the authenticity of an image by relying ...

AI Technical Summary

Problems solved by technology

[0003] Although some researchers currently propose some solutions for the synthetic face of a specific technology, such as Face2Face face synthesis technology, some researchers propose wavelet transform statistical moment features or SRM residual features to describe natural real images and The difference between synthetic images, but the detection results are not stable, and images are often transmitted in compressed form in multimedia. For compressed images, the detection performance based on these feature schemes drops significantly; another example is for the popular GAN people. Face generation technology, researchers use the color mismatch characteristics of real natural faces and generated faces in the three color spaces of RGB, HSV, and YCbCr, extract co-occurrence matrices as features to distinguish these two types of faces, or use some target recognition neural Networks like Resnet, Xception, etc. to distinguish real faces from fake faces

However, in these methods, the traditional image statistical moment features can only be detected for images generated by specific image synthesis techniques. The general neural network model is huge, the network structure is complex, the training is difficult and the training time is long.

These methods are difficult to meet the versatility and efficiency required for synthetic face forensics.

Images