适用于图像检索的强化对抗生成哈希方法  

Strengthened Generative Adversarial Hashing for Image Retrieval

作　　者：施鸿源 宋井宽 全成斌[1] 赵有健[1] SHI Hong-yuan;SONG Jing-kuan;QUAN Cheng-bin;ZHAO You-jian(Department of Computer Science and Technology,Tsinghua University,Beijing 100084,China;Center for Future Media and School of Computer Science and Engineering,University of Electronic Science and Technology,Chengdu 611731,China)

机构地区：[1]清华大学计算机科学与技术系,北京100084 [2]电子科技大学未来媒体中心,成都611731

出　　处：《小型微型计算机系统》2021年第5期1039-1043,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61825204)资助.

摘　　要：在互联网上,大量的数据是由音视频等多媒体流量构成,特别是图像和视频占了绝大部分.由于视频流量检索可以转换成图像的检索,因此如何在互联网上海量数据中进行高效图像检索成了一个重要的研究领域.深度哈希方法在图像检索中可以有效提高检索速度和检索效率,故其在图像检索领域占据了重要的地位.针对大量图像数据无标注的特点,本文提出了适用于图像检索的强化对抗生成哈希方法,该方法结合最新的生成对抗网络,引入了相似度矩阵来提高算法性能,具有相比其他无监督哈希方法准确率较高的特点.文章中提出的方法在不同数据集上进行了实验,其表现都是良好的.On the Internet,most of data is made up of multimedia traffic such as audio and video.Because video retrieve can be converted into image retrieval,how to effectively in the big-data image retrieval has become an important field of research.With the deep hashing can effectively improve the retrieval speed in image retrieval,it occupies the important position in the field of image retrieval.In view of the unlabeled feature of a large number of image data,this paper proposes strengthened Generative Adversarial hashing for image retrieval,which combines with the latest generative adversarial networks and introduces the similarity matrix to improve the performance of the algorithm.The method has been tested on CIFAR-10 and NUS-WIDE,its results are good.

关 键 词：深度学习 哈希 图像检索 无监督学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]