纺织学报 ›› 2019, Vol. 40 ›› Issue (04): 117-121.doi: 10.13475/j.fzxb.20180603205

• 服装工程 • 上一篇    下一篇

采用卷积神经网络 CaffeNet 模型的女裤廓形分类

吴欢1, 丁笑君1,2, 李秦曼1, 杜磊1,2, 邹奉元1,2()   

  1. 1.浙江理工大学 服装学院, 浙江 杭州 310018
    2.浙江理工大学 浙江省服装工程技术研究中心, 浙江 杭州 310018
  • 收稿日期:2018-06-07 修回日期:2019-01-08 出版日期:2019-04-15 发布日期:2019-04-16
  • 通讯作者: 邹奉元
  • 作者简介:吴欢(1994—),女,硕士生。主要研究方向为服装数字化技术。
  • 基金资助:
    “浙江省服装工程技术研究中心”省部级重点实验室开放基金项目(2018FZKF13);2018年浙江省大学生科技创新活动计划项目(2018R406076)

Classification of women’s trousers silhouette using convolution neural network CaffeNet model

WU Huan1, DING Xiaojun1,2, LI Qinman1, DU Lei1,2, ZOU Fengyuan1,2()   

  1. 1. School of Fashion Design & Engineering, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China
    2. Zhejiang Provincial Research Center of Clothing Engineering Technology,Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China
  • Received:2018-06-07 Revised:2019-01-08 Online:2019-04-15 Published:2019-04-16
  • Contact: ZOU Fengyuan

摘要:

针对服装廓形分类特征提取计算复杂、分类效果尚不理想等问题,提出了一种基于卷积神经网络CaffeNet模型的服装廓形分类方法。以女裤为例,首先建立一个包括吊裆裤、阔腿裤、喇叭裤、小脚裤和直筒裤的5种女裤廓形样本库,利用卷积神经网络相互交替的卷积层和池化层从女裤图像中自动提取形状特征,通过反向传播算法不断逐层更新权值,采用梯度下降法并且改进全连接层的参数最小化损失函数,运用Softmax回归分类器来实现女裤的廓形分类。实验结果表明,该方法能够有效地对女裤廓形进行分类,分类准确率达到95%以上,可为服装商品的可视化分类识别提供有效途径。

关键词: 卷积神经网络, CaffeNet模型, 女裤廓形, Softmax回归

Abstract:

Aiming at the complicated calculation of clothing silhouette classification feature extraction and poor classification effect, a classification approach of clothing silhouette based on the CaffeNet model of convolution neural network was proposed. Taking women’s trousers as an example, a sample database of five kinds of women’s trousers with silhouette was established at first, comprising saggy pants, broad-legged pants, flared trousers, pencil pants and straight pants, then shape features were extracted automatically from the clothing images using the alternating convolution and pool layers, weight values were updated by back propagation algorithm layer by layer, the gradient descent method was adopted and the parameter of the whole connection layer was modified to minimize loss function, and Softmax regression was used to classify the women’s trousers silhouette. The experimental results show that the novel approach can classify the silhouette of women’s trousers accurately, and the classification accuracy is up to 95%. It can provide an effective way for visual classification and recognition of clothing products.

Key words: convolution neural network, CaffeNet model, women's trousers silhouette, Softmax regression

中图分类号: 

  • TS941.26

图1

5种女裤的样本"

图2

卷积神经网络结构"

表1

网络的隐含层参数"

层数 每层类型 卷积核大小(个数) 步长
1 卷积层1 11×11核(96个) 4
2 池化层1 3×3核(1个) 2
3 卷积层2 5×5核(256个) 1
4 池化层2 3×3核(1个) 2
5 卷积层3 3×3核(384个) 1
6 卷积层4 3×3核(384个) 1
7 卷积层5 3×3核(256个) 1
8 池化层5 3×3核(1个) 2
9 全连接层6 1×1核(1个) 1
10 全连接层7 1×1核(1个) 1
11 全连接层8 1×1核(1个) 1

图3

图像预处理过程"

表2

改进的CNN 对廓形分类的准确率"

女裤种类 训练集数量/张 测试集数量/张 准确率/%
吊裆裤 240 60 95.0
阔腿裤 240 60 96.7
喇叭裤 240 60 96.7
小脚裤 240 60 98.3
直筒裤 240 60 95.0

表3

FD + SVM 对廓形分类的准确率"

女裤种类 训练集数量/张 测试集数量/张 准确率/%
吊裆裤 240 60 85.0
阔腿裤 240 60 90.0
喇叭裤 240 60 88.3
小脚裤 240 60 88.3
直筒裤 240 60 86.7
[1] BOSSARD L, DANTONE M, LEISTNER C, et al. Apparel classification with style[C]// Asian Conference on Computer Vision. Korea: Springer-Verlag Vision, 2012: 321-335.
[2] DI W, WAH C, BHARDWAJ A, et al. Style finder: fine-grained clothing style recognition and retri-eval[C]// IEEE Conference on Computer Vision and Pattern Recognition. Portland: IEEE Computer Society Foundation, 2013: 8-10.
[3] ZHANG W, ANTUNEZ E, GOKTURK S, et al. Apparel silhouette attributes recognition [C]//IEEE Workshop on the Applications of Computer Vision. Breckenridge: IEEE Computer Society, 2012: 489-496.
[4] AN L X, LI W. An integrated approach to fashion flat sketches classification[J]. International Journal of Clothing Science and Technology, 2014,26(5):346-366.
[5] 李东, 万贤福, 汪军. 采用傅里叶描述子和支持向量机的服装款式识别方法[J]. 纺织学报, 2017,38(5):122-127.
LI Dong, WAN Xianfu, WANG Jun. Clothing style recognition approach using fourier descriptors and support vector machines[J]. Journal of Textile Research, 2017,38(5):122-127.
[6] YAMAGUCHI K, KIAPOUR H, ORTIZ L, et al. Parsing clothing in fashion photographs[C]// IEEE Conference on Computer Vision and Pattern Recognition. Providence: IEEE Computer Science, 2012: 3570-3577.
[7] ZAGORUYKO S, KOMODAKIS N. Learning to compare image patches via convolutional neural networks[C]// IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE Computer Society, 2015: 4353-4361.
[8] JIA Y, SHELHAMER E, DONAHUE J. Caffe: convolutional architecture for fast feature embedding[C]// 22nd ACM International Conference on Multimedia. Orlando: Computer Science, 2014: 675-678.
[9] KARPATHY A, TODERICI G, SHETTY S, et al. Large-scale video classification with convolutional neural networks[C]// IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE Computer Society, 2014: 1725-1732.
[10] SERMANET P, LECUN Y. Traffic sign recognition with multi-scale convolutional networks[C]// International Joint Conference on Neural Networks. California: International Neural Network Society, 2011: 2809-2813.
[11] CUI X, LIU Y, WANG C, et al. Defect classification for tire X-ray images using convolutional neural network[J]. Electronic Measurement Technology, 2017,40(5):168-173.
[12] CHANG C C. LIN C J, LIBSVM: A library for support vector machines [CP/OL]. 2010-03-11 [2015-11-09]. http://www.csie.ntu.edu.tw/~cjlin/libsvm.
[1] 邵金鑫, 张宝昌, 曹继鹏. 基于图像处理与深度学习方法的棉纤维梳理过程纤维检测识别技术[J]. 纺织学报, 2020, 41(07): 40-46.
[2] 王泽霞, 陈革, 陈振中. 基于改进卷积神经网络的化纤丝饼表面缺陷识别[J]. 纺织学报, 2020, 41(04): 39-44.
[3] 王晓华, 姚炜铭, 王文杰, 张蕾, 李鹏飞. 基于改进YOLO 深度卷积神经网络的缝纫手势检测[J]. 纺织学报, 2020, 41(04): 142-149.
[4] 贾小军, 叶利华, 邓洪涛, 刘子豪, 陆锋杰. 基于卷积神经网络的蓝印花布纹样基元分类 [J]. 纺织学报, 2020, 41(01): 110-117.
[5] 孙洁, 丁笑君, 杜磊, 李秦曼, 邹奉元. 基于卷积神经网络的织物图像特征提取与检索研究进展[J]. 纺织学报, 2019, 40(12): 146-151.
[6] 刘正东, 刘以涵, 王首人. 西装识别的深度学习方法[J]. 纺织学报, 2019, 40(04): 158-164.
[7] 汪珊娜 张华熊 康锋. 基于卷积神经网络的领带花型情感分类[J]. 纺织学报, 2018, 39(08): 117-123.
[8] 王雯雯 高畅 刘基宏. 应用卷积神经网络的细纱断纱锭位识别[J]. 纺织学报, 2018, 39(06): 136-141.
[9] 景军锋 范晓婷 李鹏飞 洪良. 应用深度卷积神经网络的色织物缺陷检测[J]. 纺织学报, 2017, 38(02): 68-74.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!