Convolutional Neural Networks for Clothes Categories
Clothes classification is a promising research topic. Due to the manually-designed features’ limitation, the existing algorithms have a problem of low accuracy in attributes classification. In this paper, we propose a new method to utilize convolutional d
- PDF / 1,124,345 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 94 Downloads / 280 Views
Abstract. Clothes classification is a promising research topic. Due to the manually-designed features’ limitation, the existing algorithms have a problem of low accuracy in attributes classification. In this paper, we propose a new method to utilize convolutional deep learning for clothes classification. We firstly set up a new database by downloading the images of each category from Internet via related software and manual work, which divides clothes into 16 categories according to the common clothing style in the market. Then, the paper designs convolutional neural networks(CNNs) architecture and adaptively learns the feature representation of clothes from our constructed dataset. The experiment adopts Bag of Words (BOW), Histogram of Oriented Gradient (HOG)+ Support Vector Machine(SVM)and HSV (Hue, Saturation, Value)+SVM to test the new database and compares these methods with our CNNs model. The results demonstrate the superiority of our CNNs to the other algorithms. Keywords: Clothes categories networks
1
·
Deep learning
·
Convolutional neural
Introduction
The 2013 annual Chinese apparel e-commerce operation report showed that variety of goods in the online market were exponentially expanding to meet customer’s increasing demands, especially clothes and footwear products. In 2013, clothes and footwear products took up the highest market share in the online market, purchasing rate reached up to 76.2%. Thus, clothes and footwear products have become the most promising goods in the online market. Nowadays, most commercial image retrieval systems mainly rely on the key words search, such as TaoBao, JingDong and SuNing e-commerce. However, these systems have two weaknesses: first, every original image needs to be marked with key word. With the widely spread of smartphones, numerous images are updated everyday. It costs a large amount of human resource and materials to mark images. Second, because of cognition subjectivity, people may have different understandings of the same image, which will result in subjectivity and inaccuracy when the images are marked by different key words. c Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part II, CCIS 547, pp. 120–129, 2015. DOI: 10.1007/978-3-662-48570-5 12
Convolutional Neural Networks for Clothes Categories
121
Many researchers have devoted to designing automatic classification of clothing. Pan et al. [1] proposed a BP neural network to recognize woven fabric. Ben et al. [2] recognized woven fabric based on the texture features and SVM classifier. Yamaguchi et al. [3] described clothes by labeling superpixels, which were obtained from image segmentation making use of a Conditional Random Field model. Liu et al. [4] had a proposal for describing clothes based on pose estimation and using the features like color, SIFT and HOG and classified clothes into 23 categories. Bourdev et. al. [5] proposed a system describe the appearance of people by using 9 binary attributes such as male/female with T-shirt and long hair. For clothes segmentation, Manfre
Data Loading...
