Energy-constrained Self-training for Unsupervised Domain Adaptation

被引:20
|
作者
Liu, Xiaofeng [1 ,2 ,7 ]
Hu, Bo [1 ,2 ,4 ]
Liu, Xiongchang [5 ]
Lu, Jun [1 ,2 ]
You, Jane [6 ]
Kong, Lingsheng [3 ]
机构
[1] Beth Israel Deaconess Med Ctr, Boston, MA 02215 USA
[2] Harvard Med Sch, Boston, MA 02115 USA
[3] Chinese Acad Sci, CAS, Changchun Inst Opt Fine Mech & Phys, Changchun, Peoples R China
[4] IIT, Chicago, IL 60616 USA
[5] China Univ Min & Technol, Dept Informat & Elect Engn, Xuzhou, Peoples R China
[6] Hong Kong Polytech Univ, Dept Comp, Hung Hom, Hong Kong, Peoples R China
[7] Fanhan Tech Inc, Suzhou, Jiangsu, Peoples R China
基金
中国国家自然科学基金;
关键词
D O I
10.1109/ICPR48806.2021.9413284
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Unsupervised domain adaptation (UDA) aims to transfer the knowledge on a labeled source domain distribution to perform well on an unlabeled target domain. Recently, the deep self-training involves an iterative process of predicting on the target domain and then taking the confident predictions as hard pseudo-labels for retraining. However, the pseudo-labels are usually unreliable, and easily leading to deviated solutions with propagated errors. In this paper, we resort to the energy-based model and constrain the training of the unlabeled target sample with the energy function minimization objective. It can be applied as a simple additional regularization. In this framework, it is possible to gain the benefits of the energy-based model, while retaining strong discriminative performance following a plug-and-play fashion. We deliver extensive experiments on the most popular and large scale UDA benchmarks of image classification as well as semantic segmentation to demonstrate its generality and effectiveness.
引用
收藏
页码:7515 / 7520
页数:6
相关论文
共 50 条
  • [1] Energy-based Self-Training and Normalization for Unsupervised Domain Adaptation
    Herath, Samitha
    Fernando, Basura
    Abbasnejad, Ehsan
    Hayat, Munawar
    Khadivi, Shahram
    Harandi, Mehrtash
    Rezatofighi, Hamid
    Haffari, Gholamreza
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 11619 - 11628
  • [2] Unsupervised Domain Adaptation with Multiple Domain Discriminators and Adaptive Self-Training
    Spadotto, Teo
    Toldo, Marco
    Michieli, Umberto
    Zanuttigh, Pietro
    [J]. 2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR), 2021, : 2845 - 2852
  • [3] DUAL-CONSISTENCY SELF-TRAINING FOR UNSUPERVISED DOMAIN ADAPTATION
    Wang, Jie
    Zhong, Chaoliang
    Feng, Cheng
    Sun, Jun
    Ide, Masaru
    Yokota, Yasuto
    [J]. 2021 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2021, : 1529 - 1533
  • [4] Contrastive Learning and Self-Training for Unsupervised Domain Adaptation in Semantic Segmentation
    Marsden, Robert A.
    Bartler, Alexander
    Doebler, Mario
    Yang, Bin
    [J]. 2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2022,
  • [5] Unsupervised Arabic Dialect Adaptation with Self-Training
    Novotney, Scott
    Schwartz, Rich
    Khudanpur, Sanjeev
    [J]. 12TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION 2011 (INTERSPEECH 2011), VOLS 1-5, 2011, : 548 - +
  • [6] Unsupervised Domain Adaptation for Medical Image Segmentation by Disentanglement Learning and Self-Training
    Xie, Qingsong
    Li, Yuexiang
    He, Nanjun
    Ning, Munan
    Ma, Kai
    Wang, Guoxing
    Lian, Yong
    Zheng, Yefeng
    [J]. IEEE TRANSACTIONS ON MEDICAL IMAGING, 2024, 43 (01) : 4 - 14
  • [7] UNSUPERVISED DOMAIN ADAPTATION FOR SPEECH RECOGNITION VIA UNCERTAINTY DRIVEN SELF-TRAINING
    Khurana, Sameer
    Moritz, Niko
    Hori, Takaaki
    Le Roux, Jonathan
    [J]. 2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021), 2021, : 6553 - 6557
  • [8] ADVERSARIAL SELF-TRAINING UNSUPERVISED DOMAIN ADAPTATION FOR REMOTE SENSING SCENE CLASSIFICATION
    Huang, Ying
    Li, Tangsheng
    Liu, Su
    Mei, Wenhao
    [J]. 2022 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM (IGARSS 2022), 2022, : 1572 - 1575
  • [9] Constraining pseudo-label in self-training unsupervised domain adaptation with energy-based model
    Kong, Lingsheng
    Hu, Bo
    Liu, Xiongchang
    Lu, Jun
    You, Jane
    Liu, Xiaofeng
    [J]. INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS, 2022, 37 (10) : 8092 - 8112
  • [10] Cycle Self-Training for Domain Adaptation
    Liu, Hong
    Wang, Jianmin
    Long, Mingsheng
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34