叶 锦,彭小江,乔 宇,邢 昊.基于深度学习的女装图片分类探索[J].集成技术,2019,8(2):1-10
基于深度学习的女装图片分类探索
Classification of Women Dress Images Based On Deep Learning
  
DOI:10.12146/j.issn.2095-3135.20181226001
中文关键词:  商品图像;卷积神经网络;采样;多任务
英文关键词:product image; convolutional neural networks; sampling; multi-task
基金项目:商品图像细粒度分类与弱监督学习算法研究项目(SY8Z003)
作者单位
叶 锦 中国科学院深圳先进技术研究院 深圳 518055 
彭小江 中国科学院深圳先进技术研究院 深圳 518055 
乔 宇 中国科学院深圳先进技术研究院 深圳 518055 
邢 昊 唯品会研究院 广州 510000 
摘要点击次数: 284
全文下载次数: 1296
中文摘要:
      互联网商品图像的属性分类是人工智能领域的重要研究课题之一,针对商品图像属性分布不 平衡以及不同属性间存在相关性等问题,该文以女装图像为分类目标,提出了一种基于卷积神经网络的商品图像分类方法。首先,从电商网站获取大量商品图像,并进行人工标注;然后,基于卷积神经 网络框架,采用了一种有效的采样策略,通过增加新的损失函数,实现了基于多任务学习方法的商品图像属性准确分类;最后,通过对不同策略下分类结果的对比分析,验证了该方法的有效性。结果显 示,所提出方法具有较高的分类精度。
英文摘要:
      With the rapid development of Internet online shopping, automatic classification of product images has become an interesting research topic. In this paper, an accurate classification method for women dress images are investigated. Firstly, 40 000 product images were crawled from the Vipshop online shopping websites, which all are annotated by several experts. Then, several baselines using deep convolutional networks were provided. Finally, a new loss function was introduced and the multi-task learning method was used to improve the classification accuracy. With the comparison of different strategies, the experimental results show that the proposed method can obtain higher classification accuracy.
查看全文  查看/发表评论  下载PDF阅读器
关闭
微信关注二维码 用微信扫一扫

美女

美女图片

美女

美女图片