Journal of Computer Applications ›› 2021, Vol. 41 ›› Issue (5): 1305-1313.DOI: 10.11772/j.issn.1001-9081.2020071059

Special Issue: 人工智能

• Artificial intelligence • Previous Articles     Next Articles

Data augmentation method based on improved deep convolutional generative adversarial networks

GAN Lan, SHEN Hongfei, WANG Yao, ZHANG Yuejin   

  1. School of Information Engineering, East China Jiaotong University, Nanchang Jiangxi 330013, China
  • Received:2020-07-20 Revised:2020-09-15 Online:2021-05-10 Published:2020-10-20
  • Supported by:
    This work is partially supported by the National Natural Science Foundation of China (11862006).


甘岚, 沈鸿飞, 王瑶, 张跃进   

  1. 华东交通大学 信息工程学院, 南昌 330013
  • 通讯作者: 沈鸿飞
  • 作者简介:甘岚(1964-),女,江西南昌人,教授,硕士,主要研究方向:模式识别、图像处理;沈鸿飞(1992-),男,安徽滁州人,硕士研究生,CCF会员,主要研究方向:图像处理;王瑶(1994-),女,安徽淮南人,硕士研究生,主要研究方向:图像识别;张跃进(1978-),男,湖北钟祥人,教授,博士,主要研究方向为:计算生物力学。
  • 基金资助:

Abstract: In order to solve the training difficulty of small sample data in deep learning and increase the training efficiency of DCGAN (Deep Convolutional Generative Adversarial Network), an improved DCGAN algorithm was proposed to perform the augmentation of small sample data. In the method, Wasserstein distance was used to replace the loss model in the original model at first. Then, spectral normalization was added in the generation network, and discrimination network to acquire a stable network structure. Finally, the optimal noise input dimension of sample was obtained by the maximum likelihood estimation and experimental estimation, so that the generated samples became more diversified. Experimental result on three datasets MNIST, CelebA and Cartoon indicated that the improved DCGAN could generate samples with higher definition and recognition rate compared to that before improvement. In particular, the average recognition rate on these datasets were improved by 8.1%, 16.4% and 16.7% respectively, and several definition evaluation indices on the datasets were increased with different degrees, suggesting that the method can realize the small sample data augmentation effectively.

Key words: small sample, data augmentation, Deep Convolutional Generative Adversarial Network (DCGAN), Wasserstein distance, spectral normalization, intrinsic dimension

摘要: 针对小样本数据在深度学习中训练难的问题,为提高DCGAN训练效率,提出了一种改进的DCGAN算法对小样本数据进行增强。首先,使用Wasserstein距离替换原模型中的损失模型;其次,在生成网络和判别网络中加入谱归一化,以得到稳定的网络结构;最后,通过极大似然估计算法和实验估算得到样本的最佳噪声输入维度,从而提高生成样本的多样性。在MNIST、CelebA和Cartoon这三个数据集上的实验结果表明:改进后的DCGAN所生成样本的清晰度以及识别率比改进前均得到了明显提高,其中平均识别率在这几个数据集上分别提高了8.1%、16.4%和16.7%,几种清晰度评价指标在各数据集上均有不同程度的提高。可见该方法能够有效地实现小样本数据增强。

关键词: 小样本, 数据增强, DCGAN, Wasserstein距离, 谱归一化, 内在维数

CLC Number: