A Review of Single-Source Deep Unsupervised Visual Domain Adaptation

被引:261
作者
Zhao, Sicheng [1 ]
Yue, Xiangyu [1 ]
Zhang, Shanghang [1 ]
Li, Bo [1 ]
Zhao, Han [2 ]
Wu, Bichen [1 ]
Krishna, Ravi [1 ]
Gonzalez, Joseph E. [1 ]
Sangiovanni-Vincentelli, Alberto L. [1 ]
Seshia, Sanjit A. [1 ]
Keutzer, Kurt [1 ]
机构
[1] Univ Calif Berkeley, Dept Elect Engn & Comp Sci, Berkeley, CA 94720 USA
[2] Carnegie Mellon Univ, Sch Comp Sci, Pittsburgh, PA 15213 USA
关键词
Task analysis; Data models; Adaptation models; Visualization; Training; Loss measurement; Learning systems; Adversarial learning; discrepancy-based methods; domain adaptation (DA); self-supervised learning (SSL); transfer learning; RECOGNITION; ROBUSTNESS; IMAGES; SHIFT;
D O I
10.1109/TNNLS.2020.3028503
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Large-scale labeled training datasets have enabled deep neural networks to excel across a wide range of benchmark vision tasks. However, in many applications, it is prohibitively expensive and time-consuming to obtain large quantities of labeled data. To cope with limited labeled training data, many have attempted to directly apply models trained on a large-scale labeled source domain to another sparsely labeled or unlabeled target domain. Unfortunately, direct transfer across domains often performs poorly due to the presence of domain shift or dataset bias. Domain adaptation (DA) is a machine learning paradigm that aims to learn a model from a source domain that can perform well on a different (but related) target domain. In this article, we review the latest single-source deep unsupervised DA methods focused on visual tasks and discuss new perspectives for future research. We begin with the definitions of different DA strategies and the descriptions of existing benchmark datasets. We then summarize and compare different categories of single-source unsupervised DA methods, including discrepancy-based methods, adversarial discriminative methods, adversarial generative methods, and self-supervision-based methods. Finally, we discuss future research directions with challenges and possible solutions.
引用
收藏
页码:473 / 493
页数:21
相关论文
共 222 条
[1]  
Achituve I., 2020, ARXIV200312641
[2]  
Adel T, 2017, AAAI CONF ARTIF INTE, P1691
[3]  
[Anonymous], 2019, Transfer Adaptation Learning: A Decade Survey
[4]   Contour Detection and Hierarchical Image Segmentation [J].
Arbelaez, Pablo ;
Maire, Michael ;
Fowlkes, Charless ;
Malik, Jitendra .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2011, 33 (05) :898-916
[5]  
Ardizzone L., 2019, P ICLR
[6]  
Athiwaratkun B., 2019, ICLR
[7]  
Azizzadenesheli K., 2019, P ICLR
[8]  
Balaji Y, 2018, ADV NEUR IN, V31
[9]  
Ballester P. L., 2019, SEM LEARN METH UNS D
[10]   A theory of learning from different domains [J].
Ben-David, Shai ;
Blitzer, John ;
Crammer, Koby ;
Kulesza, Alex ;
Pereira, Fernando ;
Vaughan, Jennifer Wortman .
MACHINE LEARNING, 2010, 79 (1-2) :151-175