Journal of Computer Applications ›› 2021, Vol. 41 ›› Issue (5): 1432-1437.DOI: 10.11772/j.issn.1001-9081.2020071138

Special Issue: 多媒体计算与计算机仿真

• Virtual reality and multimedia computing • Previous Articles     Next Articles

Image generation based on conditional-Wassertein generative adversarial network

GUO Maozu1,2, YANG Qiannan1,2, ZHAO Lingling3   

  1. 1. School of Electrical and Information Engineering, Beijing University of Civil Engineering and Architecture, Beijing 100044, China;
    2. Beijing Key Laboratory of Intelligent Processing for Building Big Data(Beijing University of Civil Engineering and Architecture), Beijing 100044, China;
    3. School of Computer Science and Technology, Harbin Institute of Technology, Harbin Heilongjiang 150001, China
  • Received:2020-07-31 Revised:2020-10-05 Online:2021-05-10 Published:2020-12-23
  • Supported by:
    This work is partially supported by Surface Program of National Natural Science Foundation of China (61871020), the Key Project of Science and Technology Plan of Beijing Municipal Education Commission (KZ201810016019), High-level Innovation Team Building Program in Beijing Municipal Colleges and Universities (IDHT20190506), the Graduate Innovation Project of Beijing University of Civil Engineering and Architecture in 2020 (PG202005).

基于条件Wassertein生成对抗网络的图像生成

郭茂祖1,2, 杨倩楠1,2, 赵玲玲3   

  1. 1. 北京建筑大学 电气与信息工程学院, 北京 100044;
    2. 建筑大数据智能处理方法研究北京市重点实验室(北京建筑大学), 北京 100044;
    3. 哈尔滨工业大学 计算机科学与技术学院, 哈尔滨 150001
  • 通讯作者: 赵玲玲
  • 作者简介:郭茂祖(1966-),男,山东德州人,教授,博士生导师,博士,主要研究方向:机器学习、智慧城市、生物信息学;杨倩楠(1995-),女,甘肃兰州人,硕士研究生,主要研究方向:机器学习、智慧城市;赵玲玲(1980-),女,黑龙江齐齐哈尔人,讲师,博士,主要研究方向:机器学习、智慧城市、生物信息学。
  • 基金资助:
    国家自然科学基金面上项目(61871020);北京市教委科技计划重点项目(KZ201810016019);北京市属高校高水平创新团队建设计划项目(IDHT20190506);北京建筑大学2020年度研究生创新项目(PG202005)。

Abstract: Generative Adversarial Network (GAN) can automatically generate target images, and is of great significance to the generation of building arrangement of similar blocks. However, there are problems in the existing process of model training such as the low accuracy of generated images, the mode collapse, and the too low efficiency of model training. To solve these problems, a Conditional-Wassertein Generative Adversarial Network (C-WGAN) model for image generation was proposed. First, the feature correspondence between the real sample and the target sample was needed to be identified by this model, and then the target sample was generated according to the identified feature correspondence. The Wassertein distance was used to measure the distance between the distributions of two image features in the model, the GAN training environment was stablized, and mode collapse was avoided during model training, so as to improve the accuracy of the generated images and the training efficiency. Experimental results show that compared with the original Conditional Generative Adversarial Network (CGAN) and the pix2pix models, the proposed model has the Peak Signal-to-Noise Ratio (PSNR) increased by 6.82% and 2.19% at most respectively; in the case of the same number of training rounds, the proposed model reaches the convergence state faster. It can be seen that the proposed model can not only effectively improve the accuracy of image generation, but also increase the convergence speed of the network.

Key words: image generation, Generative Adversarial Network (GAN), Conditional Generative Adversarial Network (CGAN), Wassertein distance

摘要: 生成对抗网络(GAN)能够自动生成目标图像,对相似地块的建筑物排布生成具有重要意义。而目前训练模型的过程中存在生成图像精度不高、模式崩溃、模型训练效率太低的问题。针对这些问题,提出了一种面向图像生成的条件Wassertein生成对抗网络(C-WGAN)模型。首先,该模型需要识别真实样本和目标样本之间特征对应关系,然后,根据所识别出的特征对应关系进行目标样本的生成。模型采用Wassertein距离来度量两个图像特征之间分布的距离,稳定GAN训练环境,规避模型训练过程中的模式崩溃,从而提升生成图像的精度和训练效率。实验结果表明,与原始条件生成对抗网络(CGAN)和pix2pix模型相比,所提模型的峰值信噪比(PSNR)分别最大提升了6.82%和2.19%;在训练轮数相同的情况下,该模型更快达到收敛状态。由此可见,所提模型不仅能够有效地提升图像生成的精度,而且能够提高网络的收敛速度。

关键词: 图像生成, 生成对抗网络, 条件生成对抗网络, Wassertein距离

CLC Number: