skip to main content
10.1145/3661638.3661657acmotherconferencesArticle/Chapter ViewAbstractPublication PagesaisnsConference Proceedingsconference-collections
research-article

Image-Category Aware Loss for Super-Multiple Category Classification on Chu Bamboo and Silk ancient Chinese characters

Published: 01 June 2024 Publication History

Abstract

In classification problems, real-world data often exhibits a long-tailed distribution: a few major categories have a large number of samples, while numerous minority categories have only a small number of samples. However, accurately identifying Chinese text images remains being a challenging task due to the vast number of classes and the characteristic long-tailed distribution observed in ancient Chinese text images in real-world scenarios. And specially, the extremely uneven category distribution deteriorates the recognition performance in the insufficient of hard and few-shot sample feature learning. To address this problem above, we devise an integrated-weighted distribution balance loss that considers both the hard and few-shot samples, that mainly induce the feature distribution and prediction ability imbalances. Firstly Image-Aware (IA) loss is proposed to adaptively reweights of hard samples, corresponding the prediction distribution imbalance issue at the image-level by assigning adaptively weights to the samples that are challenging to accurately identification. And then Category-Aware (CA) loss is devised to adaptively rebalance of weights, tackling the long-tail uneven sample distribution issue of the dataset at the category level by readjusting weights and decision boundary for tail samples. Experimental results from the integrated Image-Category Aware (ICA) loss demonstrate a significant and effective performance improvement trained on many network models with this integrated-loss function on the dataset of the Chu Bamboo and Silk 730 (Chu730) compared to existing methods.

References

[1]
JALALI A, KAVURI S, LEE M. Low-shot transfer with attention for highly imbalanced cursive character recognition [J]. Neural Networks, 2021, 143: 489-99.
[2]
WANG K, YI Y, TANG Z, Multi-scene ancient Chinese text recognition with deep coupled alignments [J]. Applied Soft Computing, 2021, 108.
[3]
NIKOLAIDOU K, SEURET M, MOKAYED H, A survey of historical document image datasets, F, 2022 [C]. Springer Science and Business Media Deutschland GmbH.
[4]
HUANG S, WANG H, LIU Y, OBC306: A large-scale oracle bone character recognition dataset; proceedings of the 15th IAPR International Conference on Document Analysis and Recognition, ICDAR 2019, September 20, 2019 - September 25, 2019, Sydney, NSW, Australia, F, 2019 [C]. IEEE Computer Society.
[5]
ZHANG Y-K, ZHANG H, LIU Y-G, Oracle character recognition by nearest neighbor classification with deep metric learning; proceedings of the 15th IAPR International Conference on Document Analysis and Recognition, ICDAR 2019, September 20, 2019 - September 25, 2019, Sydney, NSW, Australia, F, 2019 [C]. IEEE Computer Society.
[6]
HAN W, REN X, LIN H, Self-supervised Learning of Orc-Bert Augmentor for Recognizing Few-Shot Oracle Characters; proceedings of the 15th Asian Conference on Computer Vision, ACCV 2020, November 30, 2020 - December 4, 2020, Virtual, Online, F, 2021 [C]. Springer Science and Business Media Deutschland GmbH.
[7]
GUO J, WANG C, ROMAN-RANGEL E, Building hierarchical representations for oracle character and sketch recognition [J]. IEEE Transactions on Image Processing, 2016, 25(1): 104-18.
[8]
ATSUMI M, KAWANO S, MORIOKA T, Deep Learning Based Ancient Asian Character Recognition; proceedings of the 2020 International Conference on Advanced Mechatronic Systems, ICAMechS 2020, December 10, 2020 - December 13, 2020, Hanoi, Viet nam, F, 2020 [C]. IEEE Computer Society.
[9]
HE H, BAI Y, GARCIA E A, ADASYN: Adaptive synthetic sampling approach for imbalanced learning; proceedings of the 2008 International Joint Conference on Neural Networks, IJCNN 2008, June 1, 2008 - June 8, 2008, Hong Kong, China, F, 2008 [C]. Institute of Electrical and Electronics Engineers Inc.
[10]
KANG B, XIE S, ROHRBACH M, DECOUPLING REPRESENTATION AND CLASSIFIER FOR LONG-TAILED RECOGNITION; proceedings of the 8th International Conference on Learning Representations, ICLR 2020, April 30, 2020, Addis Ababa, Ethiopia, F, 2020 [C]. International Conference on Learning Representations, ICLR.
[11]
KHAN S H, HAYAT M, BENNAMOUN M, Cost-sensitive learning of deep feature representations from imbalanced data [J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(8): 3573-87.
[12]
CAO K, WEI C, GAIDON A, Learning imbalanced datasets with label-distribution-aware margin loss; proceedings of the 33rd Annual Conference on Neural Information Processing Systems, NeurIPS 2019, December 8, 2019 - December 14, 2019, Vancouver, BC, Canada, F, 2019 [C]. Neural information processing systems foundation.
[13]
LIN T-Y, GOYAL P, GIRSHICK R, Focal Loss for Dense Object Detection [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2): 318-27.
[14]
REN M, ZENG W, YANG B, Learning to reweight examples for robust deep learning [Z]. arXiv. 2018.
[15]
CUI Y, JIA M, LIN T-Y, Class-balanced loss based on effective number of samples [Z]. arXiv. 2019.
[16]
LIU Z, MIAO Z, ZHAN X, Large-scale long-tailed recognition in an open World [Z]. arXiv. 2019.
[17]
WANG Y-X, RAMANAN D, HEBERT M. Learning to model the tail; proceedings of the 31st Annual Conference on Neural Information Processing Systems, NIPS 2017, December 4, 2017 - December 9, 2017, Long Beach, CA, United states, F, 2017 [C]. Neural information processing systems foundation.
[18]
TAN J, WANG C, LI B, Equalization loss for long-tailed object recognition; proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020, June 14, 2020 - June 19, 2020, Virtual, Online, United states, F, 2020 [C]. IEEE Computer Society.

Index Terms

  1. Image-Category Aware Loss for Super-Multiple Category Classification on Chu Bamboo and Silk ancient Chinese characters

    Recommendations

    Comments

    Information & Contributors

    Information

    Published In

    cover image ACM Other conferences
    AISNS '23: Proceedings of the 2023 International Conference on Artificial Intelligence, Systems and Network Security
    December 2023
    467 pages
    ISBN:9798400716966
    DOI:10.1145/3661638
    Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected].

    Publisher

    Association for Computing Machinery

    New York, NY, United States

    Publication History

    Published: 01 June 2024

    Permissions

    Request permissions for this article.

    Check for updates

    Qualifiers

    • Research-article
    • Research
    • Refereed limited

    Conference

    AISNS 2023

    Contributors

    Other Metrics

    Bibliometrics & Citations

    Bibliometrics

    Article Metrics

    • 0
      Total Citations
    • 14
      Total Downloads
    • Downloads (Last 12 months)14
    • Downloads (Last 6 weeks)4
    Reflects downloads up to 14 Feb 2025

    Other Metrics

    Citations

    View Options

    Login options

    View options

    PDF

    View or Download as a PDF file.

    PDF

    eReader

    View online with eReader.

    eReader

    HTML Format

    View this article in HTML Format.

    HTML Format

    Figures

    Tables

    Media

    Share

    Share

    Share this Publication link

    Share on social media