본문 바로가기
  • Home

Improving Character Recognition by Using Metadata on Personal Attributes

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(2), pp.45-51
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : January 3, 2025
  • Accepted : February 21, 2025
  • Published : February 28, 2025

Jae-Guk Shin 1 Min-Seok Kim 1 Seung-Bo Park 1

1인하대학교

Accredited

ABSTRACT

This study proposes a metadata-based approach to enhance the accuracy of character recognition through facial detection and similarity measurement. Collecting attribute metadata such as gender, race, and headwear status, and adjusting the deep learning model’s weights using a proposed weight adjustment formula, overcome the limitations of existing methods and enhance recognition reliability across diverse environments. Experimental results demonstrate that the approach achieves high accuracy even in constrained environments and proves the significance of weight adjustments through metadata in enhancing character recognition performance. Future research will focus on expanding data diversity and developing robust models resistant to environmental factors like lighting changes to achieve greater reliability and generalization.

Citation status

* References for papers published after 2023 are currently being built.

This paper was written with support from the National Research Foundation of Korea.