Generalized fake face detection method based on meta-learning

Abstract

The invention discloses a generalized fake face detection method based on meta-learning, and relates to fake face detection. In order to solve the defect that a false face of an unknown attack algorithm cannot be well detected by a traditional dichotomy model-based false face detection method, different contributions of false face samples to model generalization and instability of a false face generation algorithm are considered, and a meta-learning-based generalized false face detection method is provided. The method comprises the following steps: 1) firstly, carrying out domain division on training sets of a plurality of attack algorithms, and randomly dividing a common training set and a meta-training set in each training stage; 2) performing feature extraction and loss function calculation on the common training set by using a convolutional neural network, and weighting each sample by using a small weight sensing network; and 3) calculating a loss function of the meta training set, updating parameters of the weight sensing network by using the gradient of the loss, and correcting the gradient of the common training set to increase the generalization of the model.

Description

technical field [0001] The present invention relates to fake human face detection, in particular to a generalized fake human face detection method based on meta-learning. Background technique [0002] With the rapid development of computer technology, face recognition technology has made great progress, especially the face recognition model based on deep learning is far more accurate than humans. At present, the face recognition system has been applied in every corner of our life. Advances in deep learning have also led to advances in other technologies, such as generative models, which are now capable of synthesizing images, music, video, and even human faces. These models have been widely used in daily life. However, these technologies also bring new challenges. For example, using GAN or Deepfake technology can generate fake faces, reduce the accuracy of face recognition, and endanger the privacy of the public. Not only that, since 2014, many open source fake face gener...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to address the shortcoming that general forged face detection methods cannot achieve good results on faces with unknown forgery methods, and consider different samples for different models with different generalization and changes between different domains, as well as the real face. Relative stability and diversity of fake faces, providing a generalized fake face detection method based on meta-learning

Images