生成对抗网络人脸生成及其检测技术研究  

Overview of Research on Face Generation and Detection Techniques in Generation Countermeasures Networks

作　　者：吴春生 佟晖[1] 范晓明 WU Chunsheng;TONG Hui;FAN Xiaoming(Beijing Police Academy,Beijing 102202,China)

机构地区：[1]北京警察学院,北京102202

出　　处：《数字通信世界》2023年第7期28-33,共6页Digital Communication World

摘　　要：随着AIGC的突破性进展,内容生成技术成为社会关注的热点。文章重点分析基于GAN的人脸生成技术及其检测方法。首先介绍GAN的原理和基本架构,然后阐述GAN在人脸生成方面的技术模式。重点对基于GAN在人脸语义生成方面的技术框架进行了综述,包括人脸语义生成发展、人脸语义生成的GAN实现。接着从多视图姿态生成、面部年龄改写、人脸的属性风格生成三个方面展开详细的阐述,并从政策法规、检测技术两个方面对伪造生成人脸图片的检测方法进行了分析。文中将检测技术分成基于深度学习、基于物理、基于生理学、基于人类视觉四个方面,最后对检测技术未来方向进行了展望。With the breakthrough development of AIGC,content generation technology has become a hot topic of social concern.This article focuses on GAN based face generation technology and its detection methods.Firstly,the principle and basic architecture of GAN are introduced,and then the technical model of GAN in face generation is described.The technical framework of facial semantic generation based on GAN is reviewed and analyzed,including the development of facial semantic generation and the implementation of GAN for facial semantic generation.Specifically,it elaborates on three aspects:multi view pose generation,facial age rewriting,and facial attribute style generation.This article systematically describes the detection methods for generating fake face images from two aspects:policies,regulations,and detection technology.In this paper,detection technology is divided into four aspects:based on deep learning,based on physics,based on physiology,and based on human vision,and the future direction of detection technology is prospected.

关 键 词：GAN 深度伪造 人脸生成 生成检测 AIGC 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]