基于注意力卷积神经网络的服装款式图廓特征识别方法  

A method for identifying clothing style outline features based on attention convolutional neural network

作　　者：白雪 曹涵颖 BAI Xue;CAO Hanying(Academy of Fine Arts,Chongqing University of Education,Chongqing 400065,China)

机构地区：[1]重庆第二师范学院美术学院,重庆400065

出　　处：《毛纺科技》2024年第7期116-121,共6页Wool Textile Journal

基　　金：重庆市教委人文社科项目(23SKGH377);重庆市教委科学技术研究计划项目(KJQN202201621);重庆市儿童研究院项目(CRIKT202019);重庆第二师范学院儿童视觉与服饰艺术设计协同创新中心项目(2021XJPT08);重庆第二师范学院校级科研项目(KY202103A);重庆市高校一流本科课程阶段性成果(sylkc2311hh)。

摘　　要：针对现有服装款式图廓特征识别方法存在特征识别技术复杂和识别精度不高的问题,提出基于注意力卷积神经网络的服装款式图廓特征识别方法。首先,采用数据增强方法对服装款式图廓类型标签进行分类;其次,通过损失函数计算并确定图廓特征分布梯度;然后,通过卷积神经网络构建特征识别模型;最后,引入注意力机制模块识别服装款式图廓特征。验证结果表明:与基于改进Resnet34和基于改进边缘检测算法的服装款式识别方法比,本文方法始终具有较高的复杂图廓识别精准度,对连衣裙款式样衣的图廓识别精准度可达99.1%,外套、裤子、短袖的款式均能达到90%以上。本文方法的识别效果精准有效,可推广于现实中服装款式图廓特征的识别。A clothing style contour feature recognition method based on attention convolutional neural network was proposed to address the problems of complex feature recognition techniques and low accuracy in existing clothing style contour feature recognition methods.Firstly,data augmentation methods was used to classify clothing style silhouette type labels;Then,the contour feature distribution gradient was calculated and determined through the loss function;After that,a feature recognition model was constructed using convolutional neural networks;Finally,the attention mechanism module was introduced to recognize the silhouette features of clothing styles.The verification results show that compared with the clothing style pattern recognition method based on improved Resnet34 and the clothing style recognition method based on improved edge detection algorithm,the proposed method always has a higher accuracy in identifying complex patterns.The accuracy of pattern recognition for dress style samples can reach 99.1%,and the styles of jackets,pants,and short sleeves can also reach over 90%.The recognition effect of the method in this article is accurate and effective,and can be extended to the recognition of clothing style contour features in reality.

关 键 词：注意力机制 卷积神经网络 服装款式 图廓特征 识别方法 

分 类 号：TP472[自动化与计算机技术]