基于多尺度transformer的伪造人脸检测方法
FAKE FACE DETECTION METHOD BASED ON MULTI-SCALE TRANSFORMER
黄继胜1
作者信息
- 1. 安徽理工大学计算机科学与工程学院 安徽 淮南 232001
- 折叠
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 transform-er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.
Abstract
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 en-hanced 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Key words
multi-scale transformer/convolutional block attention module/fake face classification/high-level semantic features引用本文复制引用
基金项目
安徽省自然科学基金(2008085MF220)
出版年
2024
NETL
NSTL
万方数据