基于多尺度transformer的伪造人脸检测方法  

FAKE FACE DETECTION METHOD BASED ON MULTI-SCALE TRANSFORMER

作　　者：黄继胜 HUANG Ji-sheng(School of Computer Science and Engineering,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽淮南232001

出　　处：《南阳理工学院学报》2024年第2期69-73,91,共6页Journal of Nanyang Institute of Technology

基　　金：安徽省自然科学基金(2008085MF220)。

摘　　要：考虑到目前大多数伪造人脸分类方法存在分类精度低、泛化能力差的问题,提出一种结合多尺度transformer和卷积块注意力模块的伪造人脸分类方法。多尺度transformer用来学习卷积层所提取特征图的高级语义特征,卷积块注意力模块用来增强此高级语义特征,使其更具有区分性,使用增强后的特征来进行分类。实验结果表明,该方法可以显著提高伪造人脸分类的准确性和泛化能力。Considering the problems of low classification accuracy and poor generalization ability in most current fake face classification methods,a fake face classification method combining multi-scale transformer and convolutional block attention module is proposed.The multi-scale transformer is used to learn the high-level semantic features of the feature map extracted by the convolutional layer,and the convolutional block attention module is used to enhance this high-level semantic feature to make it more discriminative,and use the enhanced features for classification.Experimental results show that the method can significantly improve the accuracy and generalization ability of fake face classification.

关 键 词：多尺度transformer 卷积块注意力模块 伪造人脸分类 高级语义特征 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]