Face image authenticity identification method based on face patch mapping

Abstract

The invention discloses a face image authenticity identification method based on face patch mapping. The face image authenticity identification method comprises the following steps: acquiring face data information, converting a single-frame image sequence from the face data information, performing face detection on the single-frame image sequence, and cutting a face region image; extracting local patches from the face region image, wherein the local patches comprise an eye eyebrow patch, a left cheek patch, a right cheek patch, a nose patch and a mouth lower jaw patch; respectively mapping the local patches to different convolutional layers of a convolutional neural network to obtain feature maps of corresponding positions and sizes; adopting a RoiAlign module to convert the feature maps with different sizes into feature maps with fixed sizes; and training a dichotomy model by using the feature map with the fixed size, and integrating dichotomy judgment results of the local patches by adopting a local voting mode to obtain a face image authenticity identification result.

Description

technical field [0001] The invention belongs to the technical field of face recognition, and in particular relates to a face image authenticity recognition method based on face patch mapping. Background technique [0002] In the past ten years, the advancement of technologies such as big data and cloud computing has provided massive data support and a wide range of application scenarios for the development of artificial intelligence. Artificial intelligence has experienced a brilliant development process. Among them, manipulating images, videos, and audio content with the help of machine learning tools, especially the "deep forgery" technology that replaces faces and reshape expressions is an important achievement in the development of artificial intelligence. Comprehensive learning of biological characteristics such as facial expressions can achieve the effect of confusing the real one, which is unmatched by any previous forgery technology. In addition, deep forgery techno...

AI Technical Summary

Problems solved by technology

[0003] With the further deepening of deep learning research in the industry, the existing forgery generator has strong adaptability, constantly self-optimizes and upgrades in the confrontation learning with the discriminator, and the generated forged images and videos are more realistic , making it difficult for the naked eye to discern its authenticity

In this case, it is very necessary to use the powerful feature expression ability of convolutional neural network (CNN) to learn the subtle discriminative information hidden in the forged data, which cannot be achieved by traditional methods.

However, most of the previous methods focus on how to build complex feature extractors to obtain global features of the full input image and dichotomy to distinguish real and fake faces, which is not optimal for ultra-realistic fakes, Because they are only slightly different, its fake images do come from real faces in some places

Images