《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (3): 764-769.DOI: 10.11772/j.issn.1001-9081.2021040788

• 2021年中国计算机学会人工智能会议(CCFAI 2021) • 上一篇    

基于单标注样本的多损失学习与联合度量视频行人重识别

殷雨昌1, 王洪元1(), 陈莉1, 冯尊登1, 肖宇2   

  1. 1.常州大学 计算机与人工智能学院 阿里云大数据学院,江苏 常州 213000
    2.常州工程职业技术学院,江苏 常州 213000
  • 收稿日期:2021-05-17 修回日期:2021-06-03 接受日期:2021-06-15 发布日期:2021-11-09 出版日期:2022-03-10
  • 通讯作者: 王洪元
  • 作者简介:殷雨昌(1996—),男,江苏盐城人,硕士研究生,主要研究方向:计算机视觉
    陈莉(1995—),女,江苏盐城人,硕士研究生,主要研究方向:计算机视觉
    冯尊登(1996—),男,安徽宿州人,硕士研究生,主要研究方向:计算机视觉
    肖宇(1981—),女,黑龙江伊春人,副教授,硕士,主要研究方向:数字媒体技术、图形图像处理。
  • 基金资助:
    国家自然科学基金资助项目(61976028)

One-shot video-based person re-identification with multi-loss learning and joint metric

Yuchang YIN1, Hongyuan WANG1(), Li CHEN1, Zundeng FENG1, Yu XIAO2   

  1. 1.School of Computer Science and Artificial Intelligence,Aliyun School of Big Data,Changzhou University,Changzhou Jiangsu 213000,China
    2.Changzhou Vocational Institute of Engineering,Changzhou Jiangsu 213000,China
  • Received:2021-05-17 Revised:2021-06-03 Accepted:2021-06-15 Online:2021-11-09 Published:2022-03-10
  • Contact: Hongyuan WANG
  • About author:YIN Yuchang, born in 1996, M. S. candidate. His research interests include computer vision.
    CHEN Li, born in 1995, M. S. candidate. Her research interests include computer vision.
    FENG Zundeng, born in 1996, M. S. candidate. His research interests include computer vision.
    XIAO Yu, born in 1981, M. S., associate professor. Her research interests include digital media technology, graphics and image processing.
  • Supported by:
    National Natural Science Foundation of China(61976028)

摘要:

为解决行人重识别标注成本巨大的问题,提出了基于单标注样本的多损失学习与联合度量视频行人重识别方法。针对标签样本数量少,得到的模型不够鲁棒的问题,提出了多损失学习(MLL)策略:在每次训练过程中,针对不同的数据,采用不同的损失函数进行优化,提高模型的判别力。其次,在标签估计时,提出了一个联合距离度量(JDM),该度量将样本距离和近邻距离结合,进一步提升伪标签预测的精度。JDM改善了无标签数据标签估计的准确率低、未标记的数据没有被充分利用导致训练过程不稳定的问题。实验结果表明,和单标注样本渐进学习方法PL相比,当每次迭代增加的伪标签样本的比率为0.10时,在MARS和 DukeMTMC-VideoReID两个数据集上的rank-1准确度达到了65.5%和76.2%,分别提升了7.6和5.2个百分点。

关键词: 视频行人重识别, 单标注样本学习, 半监督学习, 标签估计, 距离度量

Abstract:

In order to solve the problem of huge labeling cost for person re-identification, a method of one-shot video-based person re-identification with multi-loss learning and joint metric was proposed. Aiming at the problem that the number of label samples is small and the model obtained is not robust enough, a Multi-Loss Learning (MLL) strategy was proposed. In each training process, different loss functions were used for different data to optimize and improve the discriminative ability of the model. Secondly, a Joint Distance Metric (JDM) was proposed for label estimation, which combined the sample distance and the nearest neighbor distance to further improve the accuracy of pseudo label prediction. JDM solved the problems of the low accuracy of label estimation for unlabeled data, and the instability in the training process caused by the unlabeled data not fully utilized. Experimental results show that compared with the one-shot progressive learning method PL (Progressive Learning), the rank-1 accuracy reaches 65.5% and 76.2% on MARS and DukeMTMC-VideoReID datasets when the ratio of pseudo label samples added per iteration is 0.10, with the improvement of the proposed method of 7.6 and 5.2 percentage points, respectively.

Key words: video-based person re-identification, one-shot learning, semi-supervised learning, label estimation, distance metric

中图分类号: