您的位置:山东大学 -> 科技期刊社 -> 《山东大学学报(工学版)》

山东大学学报 (工学版) ›› 2019, Vol. 49 ›› Issue (2): 61-66.doi: 10.6040/j.issn.1672-3961.0.2017.432

• 机器学习与数据挖掘 • 上一篇    下一篇

基于深度卷积神经网络的图像自编码算法

何奕江1(),杜军平1,*(),寇菲菲1,梁美玉1,王巍2,罗盎2   

  1. 1. 北京邮电大学计算机学院, 北京 100876
    2. 新浪技术(中国)有限公司, 北京 100876
  • 收稿日期:2017-05-05 出版日期:2019-04-20 发布日期:2019-04-19
  • 通讯作者: 杜军平 E-mail:he66024748@163.com;junpingdu@126.com
  • 作者简介:何奕江(1994—),男,湖南道县人,硕士,主要研究方向为机器学习与数据挖掘. E-mail:he66024748@163.com
  • 基金资助:
    国家自然科学基金重点项目(61532006);国家自然科学基金国际合作项目(61320106006);国家自然科学基金青年科学基金(61502042)

Images auto-encoding algorithm based on deep convolution neural network

Yijiang HE1(),Junping DU1,*(),Feifei KOU1,Meiyu LIANG1,Wei WANG2,Ang LUO2   

  1. 1. School of Computer Science, Beijing University of Posts and Telecommunications, Beijing 100876, China
    2. Sina. Com Technology (China) Corporation, Beijing 100876, China
  • Received:2017-05-05 Online:2019-04-20 Published:2019-04-19
  • Contact: Junping DU E-mail:he66024748@163.com;junpingdu@126.com
  • Supported by:
    国家自然科学基金重点项目(61532006);国家自然科学基金国际合作项目(61320106006);国家自然科学基金青年科学基金(61502042)

摘要:

针对目前图像编码的研究工作更加重视信息无损性,而没有体现出社交网络图像区分度的问题,本研究提出一种新颖的基于深度卷积神经网络的社交网络图像自编码算法,将深度卷积神经网络提取特征的能力与社交网络中图像的特点相结合,得到性能良好的图像自编码。结合社交网络图片的特性与聚类算法,先将图片进行聚类得到距离信息,再利用深度卷积神经网络学习图片的距离信息,提取深度卷积神经网络中的全连接层作为编码,重复以上步骤,并得到最终的图像编码。试验结果表明,本研究提出的算法在图像搜索中的效果好于其他算法,更利于在社交网络图像搜索中使用。

关键词: 深度卷积, 神经网络, 社交网络图片, 图像自编码, 图像搜索

Abstract:

At present, image coding research was focused on information lossless, but it did not reflect the social network image differentiation. A novel social network images auto-encoding algorithm based on deep convolution neural network was proposed. The algorithm obtained good performance on image auto-encoding, which combined the feature extraction ability of deep convolutional neural network and characteristics of images in social networks. It combined the characteristics of the social network image with the clustering algorithm to cluster social network image and got the distance information, next the deep convolutional neural network was used to learn the distance information of these images, then it extracted the fully connected layer in the deep convolution neural network as the image coding, repeated the above steps and got the image coding finally. The experimental results showed that the proposed algorithm performed better than other algorithms of image search, and was more adaptive in the social network image search than that of the other algorithms mentioned.

Key words: deep convolution, neural network, social network pictures, image auto-encoding, image search

中图分类号: 

  • TP37

图1

基于深度卷积神经网络社交网络图像的自编码算法总体框架"

图2

距离信息提取"

图3

深度卷积神经网络编码学习与生成过程"

表1

深度卷积神经网络构架"

层数 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
功能 Input Conv3-64 Conv3-64 Maxpool Conv3-128 Conv3-128 Maxpool Conv3-256 Conv3-256 Conv3-256 Maxpool Conv3-512 Conv3-512 Conv3-512 Maxpool Conv3-512 Conv3-512 Conv3-512 FC-4 096 FC-2 048 FC-512 Softmax

表2

不同算法在社交网络图像上的前5个搜索结果的准确率"

算法 16位 32位 64位
DA 0.538 0.532 0.584
AEVB 0.554 0.422 0.356
DCNNSE-1 0.330 0.566 0.590
DCNNSE-2 0.572 0.606 0.596

图4

不同算法在社交网络图像上的前5个搜索结果的准确率柱状图"

表3

不同算法在社交网络图像上的平均准确率均值指标对比"

算法 16位 32位 64位
DA 0.180 0.202 0.218
AEVB 0.200 0.176 0.162
DCNNSE-1 0.138 0.204 0.209
DCNNSE-2 0.212 0.244 0.212

图5

不同算法在社交网络图像上的平均准确率均值指标柱状图"

1 谢易道.大规模人脸图像编码及其在人脸验证中的应用研究[D].成都:电子科技大学, 2015.
XIE Yidao. Large-scale face image coding and application in face verification[D]. Chengdu: University of Electronic Science and Technology of China. 2015.
2 张峰.基于Context建模的熵编码在无失真图像压缩中的应用[D].昆明:云南大学, 2015.
ZHANG Feng. Context-modeling based application in the lossless image compression of entropy coding[D]. Kunming: Yunnan University, 2015.
3 BENGIO Y , COURVILLE A , VINCENT P . Representation learning: a review and new perspectives[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2013, 35 (8): 1798- 1828.
4 YANG J, PARIKH D, BATRA D. Joint unsupervised learning of deep representations and image clusters[C]// IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE, 2016: 5147-5156.
5 DOERSCH C, GUPTA A, EFROS A A. Unsupervised visual representation learning by context prediction[C]// IEEE International Conference on Computer Vision. Santiago, Chile: IEEE, 2015: 1422-1430.
6 SEIFFERT U . ANNIE——artificial neural network-based image encoder[J]. Neurocomputing, 2014, 125 (3): 229- 235.
7 FELZENSZWALB P F , GIRSHICK R B , MCALLESTER D , et al. Object detection with discriminatively trained part-based models[J]. IEEE Transactions on Software Engineering, 2014, 32 (9): 1627- 45.
8 LIU L, SHEN C, WANG L, et al. Encoding high dimensional local features by sparse coding based fisher vectors[EB/OL].(2014-11)[2017-01-27].https://arxiv.org/pdf/1411.406.pdf.
9 BADRINARAYANAN V , KENDALL A , CIPOLLA R . Segnet: a deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (12): 2481- 2495.
doi: 10.1109/TPAMI.34
10 GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[EB/OL].(2013-11)[2017-01-24]. https://arxiv.org/pdf/1311.2524v3.pdf.
11 MATTAR, MOHAMED Marwan. Unsupervised joint alignment, clustering and feature learning[EB/OL]. (2014-05)[2017-01-20]. https://scholarworks.umass.edu/cgi/viewcontent.cgi?article=1121&context=dissertations_2.
12 TIAN D P . A review on image feature extraction and representation techniques[J]. International Journal of Multimedia & Ubiquitous Engineering, 2013, 8 (4): 385- 395.
13 DENNIS J , TRAN H D , CHNG E S . Image feature representation of the subband power distribution for robust sound event classification[J]. IEEE Transactions on Audio Speech & Language Processing, 2011, 21 (2): 2437- 2440.
14 JIA Y, HUANG C. Receptive field learning for pooled image features. 8781218[P]. 2014-07-15.
15 JIN X , CAI Z X . A global image feature construction method based on local jet structure[J]. Acta Automatica Sinica, 2014, 40 (6): 1148- 1155.
doi: 10.1016/S1874-1029(14)60012-4
16 SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL]. (2014-09-04)[2017-02-16].https://arxiv.org/pdf/1409.1556.pdf.
17 Kingma D P, Welling M. Auto-Encoding Variational Bayes[EB/OL].(2013-12-20)[2017-01-12].https://arxiv.org/pdf/1312.6114.pdf.
18 HEYMANN J, HAEB-UMBACH R, GOLIK P, et al. Unsupervised adaptation of a denoising autoencoder by bayesian feature enhancement for reverberant asr under mismatch conditions[C]// IEEE International Conference on Acoustics, Speech and Signal Processing. South Brisbane, Queensland, Australia: IEEE, 2015: 5053-5057.
[1] 侯霄雄,许新征,朱炯,郭燕燕. 基于AlexNet和集成分类器的乳腺癌计算机辅助诊断方法[J]. 山东大学学报 (工学版), 2019, 49(2): 74-79.
[2] 郭芳,陈蕾,杨子文. 基于MGU的大规模IP骨干网络实时流量预测[J]. 山东大学学报 (工学版), 2019, 49(2): 88-95.
[3] 杨亚楠,夏斌,谢楠,袁文浩. 基于BP神经网络和多元Taylor级数的混合定位算法[J]. 山东大学学报 (工学版), 2019, 49(1): 36-40.
[4] 权稳稳,林明星. CNN特征与BOF相融合的水下目标识别算法[J]. 山东大学学报 (工学版), 2019, 49(1): 107-113.
[5] 刘振丙,方旭升,杨辉华,蓝如师. 基于多尺度残差神经网络的阿尔茨海默病诊断分类[J]. 山东大学学报 (工学版), 2018, 48(6): 1-7, 18.
[6] 宋琦悦,穆学文,程欢. 改进滴水算法的黏连字符分割方法[J]. 山东大学学报 (工学版), 2018, 48(6): 89-94, 108.
[7] 沈冬冬,周风余,栗梦媛,王淑倩,郭仁和. 基于集成深度神经网络的室内无线定位[J]. 山东大学学报 (工学版), 2018, 48(5): 95-102.
[8] 张璞,刘畅,王永. 基于特征融合和集成学习的建议语句分类模型[J]. 山东大学学报 (工学版), 2018, 48(5): 47-54.
[9] 梁蒙蒙,周涛,夏勇,张飞飞,杨健. 基于PSO-ConvK卷积神经网络的肺部肿瘤图像识别[J]. 山东大学学报 (工学版), 2018, 48(5): 77-84.
[10] 张宪红,张春蕊. 基于六维前馈神经网络模型的图像增强算法[J]. 山东大学学报(工学版), 2018, 48(4): 10-19.
[11] 赵彦霞, 王熙照. 基于SVD和DCNN的彩色图像多功能零水印算法[J]. 山东大学学报(工学版), 2018, 48(3): 25-33.
[12] 曹雅,邓赵红,王士同. 基于单调约束的径向基函数神经网络模型[J]. 山东大学学报(工学版), 2018, 48(3): 127-133.
[13] 谢志峰,吴佳萍,马利庄. 基于卷积神经网络的中文财经新闻分类方法[J]. 山东大学学报(工学版), 2018, 48(3): 34-39.
[14] 何正义,曾宪华,郭姜. 一种集成卷积神经网络和深信网的步态识别与模拟方法[J]. 山东大学学报(工学版), 2018, 48(3): 88-95.
[15] 唐乐爽,田国会,黄彬. 一种基于DSmT推理的物品融合识别算法[J]. 山东大学学报(工学版), 2018, 48(1): 50-56.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 王素玉,艾兴,赵军,李作丽,刘增文 . 高速立铣3Cr2Mo模具钢切削力建模及预测[J]. 山东大学学报(工学版), 2006, 36(1): 1 -5 .
[2] 孔祥臻,刘延俊,王勇,赵秀华 . 气动比例阀的死区补偿与仿真[J]. 山东大学学报(工学版), 2006, 36(1): 99 -102 .
[3] 来翔 . 用胞映射方法讨论一类MKdV方程[J]. 山东大学学报(工学版), 2006, 36(1): 87 -92 .
[4] 李梁,罗奇鸣,陈恩红. 对象级搜索中基于图的对象排序模型(英文)[J]. 山东大学学报(工学版), 2009, 39(1): 15 -21 .
[5] 陈瑞,李红伟,田靖. 磁极数对径向磁轴承承载力的影响[J]. 山东大学学报(工学版), 2018, 48(2): 81 -85 .
[6] 王波,王宁生 . 机电装配体拆卸序列的自动生成及组合优化[J]. 山东大学学报(工学版), 2006, 36(2): 52 -57 .
[7] 李可,刘常春,李同磊 . 一种改进的最大互信息医学图像配准算法[J]. 山东大学学报(工学版), 2006, 36(2): 107 -110 .
[8] 季涛,高旭,孙同景,薛永端,徐丙垠 . 铁路10 kV自闭/贯通线路故障行波特征分析[J]. 山东大学学报(工学版), 2006, 36(2): 111 -116 .
[9] 秦通,孙丰荣*,王丽梅,王庆浩,李新彩. 基于极大圆盘引导的形状插值实现三维表面重建[J]. 山东大学学报(工学版), 2010, 40(3): 1 -5 .
[10] 刘文亮,朱维红,陈涤,张泓泉. 基于雷达图像的运动目标形态检测及跟踪技术[J]. 山东大学学报(工学版), 2010, 40(3): 31 -36 .