基于半监督深度生成对抗网络的图像识别方法  

Image Recognition Based on Semi Supervised Deep Convolutional Generative Adversarial Networks

作　　者：曾琦 向德华 李宁 肖红光[1] ZENG Qi;XIANG De-hua;LI Ning;XIAO Hong-guang(School of Computer&Communication Engineering,Changsha University of Science&Technology,Changsha 410114,China;Hunan Institute of Metrology and Test,Changsha 410014,China)

机构地区：[1]长沙理工大学计算机与通信工程学院,湖南长沙410114 [2]湖南省计量检测研究院,湖南长沙410014

出　　处：《测控技术》2019年第8期37-42,共6页Measurement & Control Technology

基　　金：国家自然科学基金青年科学基金项目(41201468);国家公益性行业科研专项(201510003-5)

摘　　要：基于生成对抗网络的图像识别方法拥有很高的识别率,但训练时需要大量有标签样本,在有标签样本较少的情况时识别效果不佳。针对这个问题,结合深度卷积生成对抗网络和半监督生成对抗网络的特点建立半监督深度生成对抗网络。根据有标签样本和无标签样本分布,模型生成拟合真实分布的样本输入并训练分类器,增加了训练样本数从而提升识别率。将模型优化调整并进行图像识别实验,结果表明,该方法仅用少量有标签样本即可实现准确的图像识别。The image recognition method based on GAN(generative adversarial networks)has a high recognition rate,but a large number of labeled samples are needed for training.Therefore,when the number of labeled samples is small,the performance of this method is poor.In order to solve this problem,a SS-DCGAN(semi supervised deep convolutional generative adversarial networks)is established,which combines the characteristics of DCGAN(deep convolutional generative adversarial networks)and SSGAN(semi supervised generative adversarial networks).According to the distribution of labeled samples and unlabeled samples,the model generates samples that fit the real data distribution and trains the classifier,and increases the number of training samples to improve the recognition rate.The model was optimized,and the experiments of image recognition were carried out.The results show that the method can realize accurate image recognition with only a small number of labeled samples.

关 键 词：生成对抗网络 半监督 深度卷积网络 图像识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]