Journal of Computer Applications ›› 2024, Vol. 44 ›› Issue (10): 2995-3010.DOI: 10.11772/j.issn.1001-9081.2023101415

• Artificial intelligence • Previous Articles     Next Articles

Overview of deep metric learning

Wenze CHAI1,2,3, Jing FAN1,2,3(), Shukui SUN1,2,3, Yiming LIANG1,2,3, Jingfeng LIU1,2,3   

  1. 1.School of Electrical and Information Technology,Yunnan Minzu University,Kunming Yunnan 650504,China
    2.Yunnan Key Laboratory of Unmanned Autonomous System (Yunnan Minzu University),Kunming Yunnan 650504,China
    3.Key Laboratory of Information and Communication Security and Disaster Recovery in Universities of Yunnan Province (Yunnan Minzu University),Kunming Yunnan 650504,China
  • Received:2023-10-19 Revised:2024-02-05 Accepted:2024-02-06 Online:2024-10-15 Published:2024-10-10
  • Contact: Jing FAN
  • About author:CHAI Wenze, born in 1998, M. S. candidate. His research interests include deep learning, image classification.
    SUN Shukui, born in 1996, M. S. candidate. His research interests include computer vision.
    LIANG Yiming, born in 1997, M. S. candidate. His research interests include natural language processing, sentiment analysis.
    LIU Jingfeng, born in 1999, M. S. candidate. His research interests include deep learning, image segmentation.
  • Supported by:
    National Natural Science Foundation of China(61540063);Youth Foundation of Humanities and Social Sciences of Ministry of Education(20YJCZH129);Project of Wu Zhonghai Expert Workstation(202305AF150045);Scientific Research Foundation of Education Department of Yunnan Province(2023Y0499);Yunnan Minzu University Master’s Research and Innovation Fund(2022SKY004)

深度度量学习综述

柴汶泽1,2,3, 范菁1,2,3(), 孙书魁1,2,3, 梁一鸣1,2,3, 刘竟锋1,2,3   

  1. 1.云南民族大学 电气信息工程学院,昆明 650504
    2.云南省无人自主系统重点实验室(云南民族大学),昆明 650504
    3.云南省高校信息与通信安全灾备重点实验室(云南民族大学),昆明 650504
  • 通讯作者: 范菁
  • 作者简介:柴汶泽(1998—),男,山西朔州人,硕士研究生,CCF会员,主要研究方向:深度学习、图像分类
    范菁(1976—),女(傣族),云南西双版纳人,教授,博士,CCF会员,主要研究方向:机器学习、模式识别、物联网 fanjing9476@ymu.edu.cn
    孙书魁(1996—),男,河南平舆人,硕士研究生,CCF会员,主要研究方向:计算机视觉
    梁一鸣(1997—),男,河南商丘人,硕士研究生,CCF会员,主要研究方向:自然语言处理、情感分析
    刘竟锋(1999—),男,湖南衡阳人,硕士研究生,主要研究方向:深度学习、图像分割。
  • 基金资助:
    国家自然科学基金资助项目(61540063);教育部人文社会科学研究青年基金资助项目(20YJCZH129);云南省吴中海专家工作站项目(202305AF150045);云南省教育厅科学研究基金资助项目(2023Y0499);云南民族大学硕士研究生科研创新基金资助项目(2022SKY004)

Abstract:

With the rise of deep neural network, Deep Metric Learning (DML) has attracted widespread attention. To gain a deeper understanding of deep metric learning, firstly, the limitations of traditional metric learning methods were organized and analyzed. Secondly, DML was discussed from three types, including types based on sample pairs, proxies, and classification. Divergence methods, ranking methods and methods based on Generative Adversarial Network (GAN) were introduced in detail of the type based on sample pairs. Proxy-based types was mainly discussed in terms of proxy samples and categories. Cross-modal metric learning, intra-class and inter-class margin problems, hypergraph classification, and combinations with other methods (such as reinforcement learning-based and adversarial learning-based methods) were discussed in the classification-based type. Thirdly, various metrics for evaluating the performance of DML were introduced, and the applications of DML in different tasks, including face recognition, image retrieval, and person re-identification, were summarized and compared. Finally, the challenges faced by DML were discussed and some possible solution strategies were proposed.

Key words: deep neural network, Deep Metric Learning (DML), machine learning, computer vision, artificial intelligence

摘要:

随着深度神经网络的兴起,深度度量学习(DML)引起广泛的关注。为了深入了解深度度量学习,首先,整理和分析传统度量学习方法的局限性。其次,从3个类型探讨DML,包括基于样本对、代理和分类的类型:基于样本对的类型包括散度方法、排序方法和基于生成对抗网络(GAN)的方法;基于代理的类型主要从代理样本、类别方面进行讨论;基于分类的类型中主要讨论了跨模态度量学习、类内类间边距问题、超图分类,以及与其他方法(如基于强化学习和基于对抗学习的方法)的结合。再次,介绍评估DML性能的各种指标,同时总结和对比DML在不同任务(包括人脸识别、图像检索和行人重识别等)中的应用。最后,探讨DML面临的挑战,并提出一些可能的解决策略。

关键词: 深度神经网络, 深度度量学习, 机器学习, 计算机视觉, 人工智能

CLC Number: