EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

WMAR-CB: Wavelet-Enhanced Multi-Strategy Network for Oracle-Bone Script Recognition

Xiang Zeng, Qiumei Peng, Yi Xu, Quanjiang Liang, Qian Yang

IEEE Access · 2025 · 13 · 210487-210496

DOI: 10.1109/access.2025.3640506 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2025. Verfasst von Xiang Zeng und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Applying computer vision to cultural heritage accelerates more efficient and intelligent scholarship, but artifacts pose unique challenges—notably model adaptation and robust feature representation. We observe that single-character oracle-bone images extracted from artifacts suffer from large-area noise that often blurs or even removes character structure, increasing recognition difficulty. Moreover, the scarcity of artifacts induces severe long-tail class distributions, biasing models toward high-frequency classes. To address these issues, we propose WMAR-CB, a recognition framework designed for noisy, single-character oracle-bone images. The model enhances both noise robustness and class balance by integrating multi-stage feature denoising and structure-aware representation learning, together with a class-balanced training objective tailored to long-tailed data. Experiments on three public datasets demonstrate substantial improvements in both overall accuracy and per-class performance. Experiments on three benchmarks demonstrate the effectiveness of WMAR-CB: on OBC-306 and Oracle-AYNU we achieve total accuracies of 92.86% and 88.36% and mean per-class accuracies of 85.39% and 83.97%, respectively; on Oracle-MNIST we reach 98.32% total accuracy.

Zitieren

APA

Zeng, X., Peng, Q., Xu, Y., Liang, Q., & Yang, Q. (2025). WMAR-CB: Wavelet-Enhanced Multi-Strategy Network for Oracle-Bone Script Recognition. IEEE Access, 13, 210487-210496. https://doi.org/10.1109/access.2025.3640506

Vancouver

Zeng X, Peng Q, Xu Y, Liang Q, Yang Q. WMAR-CB: Wavelet-Enhanced Multi-Strategy Network for Oracle-Bone Script Recognition. IEEE Access. 2025;13:210487-210496. doi:10.1109/access.2025.3640506

BibTeX
@article{Zeng2025,
  author = {Zeng, Xiang and Peng, Qiumei and Xu, Yi and Liang, Quanjiang and Yang, Qian},
  title = {WMAR-CB: Wavelet-Enhanced Multi-Strategy Network for Oracle-Bone Script Recognition},
  journal = {IEEE Access},
  year = {2025},
  volume = {13},
  pages = {210487--210496},
  doi = {10.1109/access.2025.3640506}
}
Kostenlos starten →