Distributionally Robust Loss for Long-Tailed Multi-label Image Classification

被引:0
|
作者
Lin, Dekun [1 ,2 ]
Peng, Tailai [1 ,2 ]
Chen, Rui [1 ,2 ]
Xie, Xinran [1 ,2 ]
Qin, Xiaolin [1 ,2 ]
Cui, Zhe [1 ,2 ]
机构
[1] Chinese Acad Sci, Chengdu Inst Comp Applicat, Chengdu 610213, Peoples R China
[2] Univ Chinese Acad Sci, Beijing 100049, Peoples R China
来源
关键词
Long-tailed learning; Multi-label classification; Loss;
D O I
10.1007/978-3-031-73414-4_24
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The binary cross-entropy (BCE) loss function is widely utilized in multi-label classification (MLC) tasks, treating each label independently. The log-sum-exp pairwise (LSEP) loss, which emphasizes higher logits for positive classes over negative ones within a sample and accounts for label dependencies, has demonstrated effectiveness for MLC. However, our experiments suggest that its performance in long-tailed multi-label classification (LTMLC) appears to be inferior to that of BCE. In this study, we investigate the impact of the log-sum-exp operation on recognition and explore optimization avenues. Our observations reveal two primary shortcomings of LSEP that lead to its poor performance in LTMLC: 1) the indiscriminate use of label dependencies without consideration of the distribution shift between training and test sets, and 2) the overconfidence in negative labels with features similar to those of positive labels. To mitigate these problems, we propose a distributionally robust loss (DR), which includes class-wise LSEP and a negative gradient constraint. Additionally, our findings indicate that the BCE-based loss is somewhat complementary to the LSEP-based loss, offering enhanced performance upon integration. Extensive experiments conducted on two LTMLC datasets, VOC-LT and COCO-LT, demonstrate the consistent effectiveness of our proposed method. Code: https://github.com/Kunmonkey/DR- Loss.
引用
收藏
页码:417 / 433
页数:17
相关论文
共 50 条
  • [1] Probability Guided Loss for Long-Tailed Multi-Label Image Classification
    Lin, Dekun
    THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 2, 2023, : 1577 - 1585
  • [2] Robust Asymmetric Loss for Multi-Label Long-Tailed Learning
    Park, Wongi
    Park, Inhyuk
    Kim, Sungeun
    Ryu, Jongbin
    2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS, ICCVW, 2023, : 2703 - 2712
  • [3] Effect of Stage Training for Long-Tailed Multi-Label Image Classification
    Yamagishi, Yosuke
    Hanaoka, Shohei
    2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS, ICCVW, 2023, : 2713 - 2720
  • [4] Does Head Label Help for Long-Tailed Multi-Label Text Classification
    Xiao, Lin
    Zhang, Xiangliang
    Jing, Liping
    Huang, Chi
    Song, Mingyang
    THIRTY-FIFTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THIRTY-THIRD CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE AND THE ELEVENTH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2021, 35 : 14103 - 14111
  • [5] Residual diverse ensemble for long-tailed multi-label text classification
    Jiangxin SHI
    Tong WEI
    Yufeng LI
    Science China(Information Sciences), 2024, 67 (11) : 92 - 105
  • [6] Residual diverse ensemble for long-tailed multi-label text classification
    Shi, Jiangxin
    Wei, Tong
    Li, Yufeng
    SCIENCE CHINA-INFORMATION SCIENCES, 2024, 67 (11)
  • [7] Exploring Contrastive Learning for Long-Tailed Multi-label Text Classification
    Audibert, Alexandre
    Gauffre, Aurelien
    Amini, Massih-Reza
    MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES: RESEARCH TRACK, PT VII, ECML PKDD 2024, 2024, 14947 : 245 - 261
  • [8] Label-Specific Feature Augmentation for Long-Tailed Multi-Label Text Classification
    Xu, Pengyu
    Xiao, Lin
    Liu, Bing
    Lu, Sijin
    Jing, Liping
    Yu, Jian
    THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 9, 2023, : 10602 - 10610
  • [9] Balancing Methods for Multi-label Text Classification with Long-Tailed Class Distribution
    Huang, Yi
    Giledereli, Buse
    Koksal, Abdullatif
    Ozgur, Arzucan
    Ozkirimli, Elif
    2021 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (EMNLP 2021), 2021, : 8153 - 8161
  • [10] Long-tailed Extreme Multi-label Text Classification by the Retrieval of Generated Pseudo Label Descriptions
    Zhang, Ruohong
    Wang, Yau-Shian
    Yang, Yiming
    Yu, Donghan
    Vu, Tom
    Lei, Likun
    17TH CONFERENCE OF THE EUROPEAN CHAPTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, EACL 2023, 2023, : 1092 - 1106