基于多特征的服装图像检索的智能购物系统研究

Research on Intelligent Shopping System Based on Multi-feature Garment Image Retrieval

现代生活是快节奏的,大多数人的生活负担过重。在这种情况下,网上购物是一个很好的节省时间模式,然而对于女士服装就不能像杂货或家具一样容易敲定。这是由于女士服装具有很多难以描述的特征,如纹理、形状、颜色、印花、长度等。对此提出了一种搜索衣服的方法,其中查询以图像的形式代替描述性集合,程序的第一步是根据衣服和袖子的长度进行识别,获得诸如颜色和纹理的下一个特征。为了检测最佳匹配,创建了1 500个图像的数据集,该数据集由craftsvilla,jabong,voonik,myntra,amazon,snapdeal,flipkart,fashionara,shoppersstop等字段构建而成。实验结果证实精确度为89.25%,召回率为87.00%。

Modern life is really fast-paced,and most people's lives are overburdened.In this case,online shopping is a good time-saving mode.However,for women's clothing,online shopping is not as easy to finalize as groceries or furniture.This is because women's clothing has many features that are difficult to describe,such as texture,shape,color,and print,length,etc.In this context,a method of searching for a garment in which the query replaces the descriptive set in the form of an image is proposed.The first step is to identify the length of the garment and the sleeve to obtain the next feature such as color and texture.In order to detect the best match,a dataset of 1500 images were created,it was constructed from fields such as craftsvilla,jabong,voonik,myntra,amazon,snapdeal,flipkart,fashionara,shoppersstop,etc.The experimental results confirmed the accuracy was 89.25%.The recall rate was 87.00%.