摘要
针对传统基于内容的服装分类对图像特征有较高的要求,当服装款式较多时,其准确率难以满足服装分类应用需求的问题,提出一种基于深度学习方法的平行自注意力分类网络.该网络在ResNet50的基础上增加了平行自注意力补偿分支,该分支能提升服装分类任务中的特征提取质量,逐步补充深层网络缺失的浅层细节信息.在数据集DeepFashion上进行了对比实验,实验结果证明了该方法的有效性.
Aiming at the problem that traditional content-based clothing classification had high requirements for image features,and its accuracy was difficult to meet the application requirements of clothing classification when there were many clothing styles,we proposed a parallel self-attention classification network based on deep learning methods.The network added a parallel self-attention compensation branch on the basis of ResNet50,which could improve the quality of feature extraction in clothing classification tasks,and gradually supplement shallow detail information missing from deep network.A comparative experiment was carried out on the DeepFashion dataset,and the experimental results proved the effectiveness of this method.
作者
朱淑畅
李文辉
ZHU Shuchang;LI Wenhui(School of Art and Design,Jilin Engineering Normal University,Changchun 130052,China;College of Computer Science and Technology,Jilin University,Changchun 130012,China)
出处
《吉林大学学报(理学版)》
CAS
北大核心
2023年第6期1419-1424,共6页
Journal of Jilin University:Science Edition
基金
吉林省科技发展计划项目(批准号:20230201082GX)。
关键词
服装类别分类
深度学习
自注意力机制
信息补偿
clothing category classification
deep learning
self-attention mechanism
information compensation