About | Contact Us | Register | Login
ProceedingsSeriesJournalsSearchEAI
Big Data Technologies and Applications. 13th EAI International Conference, BDTA 2023, Edinburgh, UK, August 23-24, 2023, Proceedings

Research Article

CTL-I: Infrared Few-Shot Learning via Omnidirectional Compatible Class-Incremental

Cite
BibTeX Plain Text
  • @INPROCEEDINGS{10.1007/978-3-031-52265-9_1,
        author={Biwen Yang and Ruiheng Zhang and Yumeng Liu and Guanyu Liu and Zhe Cao and Zhidong Yang and Heng Yu and Lixin Xu},
        title={CTL-I: Infrared Few-Shot Learning via Omnidirectional Compatible Class-Incremental},
        proceedings={Big Data Technologies and Applications. 13th EAI International Conference, BDTA 2023, Edinburgh, UK, August 23-24, 2023, Proceedings},
        proceedings_a={BDTA},
        year={2024},
        month={1},
        keywords={Infrared Few-shot Learning Class-incremental Learning},
        doi={10.1007/978-3-031-52265-9_1}
    }
    
  • Biwen Yang
    Ruiheng Zhang
    Yumeng Liu
    Guanyu Liu
    Zhe Cao
    Zhidong Yang
    Heng Yu
    Lixin Xu
    Year: 2024
    CTL-I: Infrared Few-Shot Learning via Omnidirectional Compatible Class-Incremental
    BDTA
    Springer
    DOI: 10.1007/978-3-031-52265-9_1
Biwen Yang1, Ruiheng Zhang1,*, Yumeng Liu2, Guanyu Liu1, Zhe Cao1, Zhidong Yang1, Heng Yu1, Lixin Xu1
  • 1: Beijing Institute of Technology, No.5 Yard
  • 2: Beijing Key Laboratory of Human-Computer Interaction, Institute of Software
*Contact email: ruiheng.zhang@bit.edu.cn

Abstract

Accommodating infrared novel class in deep learning models without sacrificing prior knowledge of base class is a challenging task , especially when the available data for the novel class is limited. Existing infrared few-shot learning methods mainly focus on measuring similarity between novel and base embedding spaces or transferring novel class features to base class feature spaces. To address this issue, we propose Infrared (omnidirectional) Compatibility Training Learning (CTL-I). We suggest building a virtual infrared prototype in the basic model to preserve feature space for potential new classes in advance. We use a method of coupling virtual and real data to gradually update these virtual prototypes as predictions for potential new categories, resulting in a more powerful classifier that can effectively adapt to new categories while retaining knowledge about general infrared features learned from the base class. Our empirical results demonstrate that our approach outperforms existing few-shot incremental learning methods on various benchmark datasets, even with extremely limited instances per class. Our work offers a promising direction for addressing the challenges of few-shot incremental learning in infrared image.

Keywords
Infrared Few-shot Learning Class-incremental Learning
Published
2024-01-31
Appears in
SpringerLink
http://dx.doi.org/10.1007/978-3-031-52265-9_1
Copyright © 2023–2025 ICST
EBSCOProQuestDBLPDOAJPortico
EAI Logo

About EAI

  • Who We Are
  • Leadership
  • Research Areas
  • Partners
  • Media Center

Community

  • Membership
  • Conference
  • Recognition
  • Sponsor Us

Publish with EAI

  • Publishing
  • Journals
  • Proceedings
  • Books
  • EUDL