基于生成对抗网络的人脸表情数据增强方法  

Data Augmentation Method Based on Generative Adversarial Networks for Facial Expression Recognition Sets

作　　者：孙晓[1,2] 丁小龙 SUN Xiao;DING Xiaolong(Institute of Emotional Computing and System Architecture,Hefei University of Technology,Hefei 230601,China;School of Computer Science and Information Engineering,Hefei University of Technology,Hefei 230601,China)

机构地区：[1]合肥工业大学情感计算与系统结构研究所,合肥230601 [2]合肥工业大学计算机与信息学院,合肥230601

出　　处：《计算机工程与应用》2020年第4期115-121,共7页Computer Engineering and Applications

基　　金：国家JW科技创新特区项目（No.JD2017XATQ0026）

摘　　要：基于深度学习的方法已经在人脸表情识别中取得了重大进展,然而人脸表情数据库的规模普遍不大。为了解决数据量不足的问题,提出了一种静态图像数据增强方法。在StarGAN的基础上修改重构误差实现多风格人脸表情图像转换,利用生成器由某一表情下的面部图像生成同一人其他表情的面部图像。在CK+表情库上的实验表明,该方法有利于提高人脸表情识别模型的识别率和泛化能力,同时对解决数据量不平衡的问题也有借鉴作用。Deep learning methods have significantly advanced in facial expression recognition.But,facial expression databases usually do not have enough data.To solve this problem,this paper proposes a static image data augmentation method.A multi-domain image-to-image translation model based on StarGAN is implemented by modifying the reconstruction loss,which can generate multi-expression facial images from the one of a certain expression.Experiments on CK+expression database show that this method can improve the accuracy and generalization capacity of recognition models,and can be used for reference to solve the problem of data imbalance.

关 键 词：数据增强 生成对抗网络 人脸表情识别 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]