《计算机应用研究》|Application Research of Computers

改进卷积神经网络在分类与推荐中的实例应用

Application of improved convolution neural network in classification and recommendation

免费全文下载 (已被下载 次)  
获取PDF全文
作者 杨天祺,黄双喜
机构 清华大学 自动化系,北京 100084
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2018)04-0974-04
DOI 10.3969/j.issn.1001-3695.2018.04.003
摘要 在网络购物不断发展的背景下,基于服装图片的服装分类识别和搭配推荐具有给予消费者搭配建议并帮助商家促进销售的重要意义。深度学习作为机器学习领域的最新研究成果,建模与表征能力强大,在图像处理领域取得了突破成果;改进卷积神经网络通过加入批量归一化、改进卷积层结构、添加冗余分类器改进了原始GoogleNet卷积神经网络,提高了分类精确度和速度。对搭配库训练集进行图片增广,扩增数据集使其更加丰富全面,并提高精确度;运用改进卷积神经网络对增广后的数据集进行服装精细分类,得到图片的服装类别风格以及功能信息;使用感知哈希算法寻找套装图片库中的相似单品及其搭配,并根据精细分类得到图片性别、风格、功能信息,最终综合给出服装搭配推荐,具有重要的现实研究意义。
关键词 服装分类与推荐;卷积神经网络;图片增广;感知哈希算法
基金项目 国家“863”计划资助项目(2015AA043702)
本文URL http://www.arocmag.com/article/01-2018-04-003.html
英文标题 Application of improved convolution neural network in classification and recommendation
作者英文名 Yang Tianqi, Huang Shuangxi
机构英文名 Dept.ofAutomation,TsinghuaUniversity,Beijing100084,China
英文摘要 In the background of the constant development of the network shopping, clothing classification and clothing collocation recommendation based on clothing pictures can provide advice to the customer and help businesses to promote sales. Deep learning is a latest research achievements in the field of machine learning, it has a strong ability of modeling and representation, and make breakthrough progress in the field of image processing. The improved convolution neural network added batch normalization structure to the network, improved the network structure of convolution layer and added redundant classifier, finally improved the classification accuracy and speed. It also augmented the images in the training set to enrich the training set, therefore improved the classification accuracy. Using the improved convolution neural network to train on the augmented training set could get the gender, style and functional information of the pictures. This paper used perceptual hash algorithm to find the similar item and its collocation suits in the repositories, together with the classification result of the gender, style and functional information of the pictures. Finally it gives the clothing collocation recommendation which has important practical research significance.
英文关键词 clothing classification and recommendation; convolution neural network; image augmentation; perceptual hash algorithm
参考文献 查看稿件参考文献
  [1] 2015H1中国服装网购行业报告[R] . 2015.
[2] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J] . Science, 2006, 313(5786):504-507.
[3] Zeiler M D, Fergus R. Visualizing and understanding convolutional networks[C] //Proc of European Conference on Computer Vision. [S. l. ] :Springer International Publishing, 2014:818-833.
[4] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J/OL] . [2015-04-10] . https://arxiv. org/abs/1409. 1556.
[5] Deng Jia, Dong Wei, Socher R, et al. ImageNet:a large-scale hierarchical image database[C] //Proc of IEEE Conference on Computer Vision and Pattern Recognition. 2009:248-255.
[6] Szegedy C, Liu Wei, Jia Yangqing, et al. Going deeper with convolutions[C] //Proc of IEEE Conference on Computer Vision and Pattern Recognition. 2015:1-9.
[7] Lin Min, Chen Qiang, Yan Shuicheng. Network in network[J/OL] . [2014-03-04] . https://arxiv. org/abs/1312. 4400.
[8] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision[C] //Proc of IEEE Conference on Computer Vision and Pattern Recognition. 2016:2818-2826.
[9] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, inception-ResNet and the impact of residual connections on learning[J/OL] . [2016-08-23] . https://arXiv. org/abs/1602. 07261.
[10] Ioffe S, Szegedy C. Batch normalization:accelerating deep network training by reducing internal covariate shift[J/OL] . [2015-03-02] . https://arxiv. org/abs/1502. 03167.
[11] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[C] //Advances in Neural Information Processing Systems. 2012:1097-1105.
[12] 刘茜, 卢心红, 李象霖. 基于多尺度 Retinex 的自适应图像增强方法[J] . 计算机应用, 2009, 29(8):2077-2079.
[13] Land E H, McCann J J. Lightness and Retinex theory[J] . Journal of the Optical Society of America, 1971, 61(1):1-11.
[14] 毛本清, 金小梅. 自适应直方图均衡化算法在图像增强处理的应用[J] . 河北北方学院学报:自然科学版, 2010, 26(5):64-68.
[15] Chetlur S, Woolley C, Vandermersch P, et al. cuDNN:efficient primitives for deep learning[J/OL] . [2014-12-18] . https://arxiv. org/abs/1410. 0759.
[16] 牛夏牧, 焦玉华. 感知哈希综述[J] . 电子学报, 2008, 36(7):1405-1411.
[17] 赵杰. 基于 DCT 均值的图像零水印算法[J] . 系统仿真技术, 2016, 11(4):304-306, 312.
[18] 张焕炯, 王国胜, 钟义信. 基于汉明距离的文本相似度计算[J] . 计算机工程与应用, 2001, 37(19):21-22.
收稿日期 2016/12/14
修回日期 2017/2/14
页码 974-977,1045
中图分类号 TP183
文献标志码 A