The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Pengesan aksara terkini telah dimodelkan menggunakan rangkaian saraf dalam dan telah mencapai prestasi tinggi dalam pelbagai tugas, seperti pengesanan teks dalam adegan semula jadi dan pengesanan watak dalam dokumen sejarah. Walau bagaimanapun, kaedah sedia ada tidak dapat mencapai ketepatan pengesanan yang tinggi untuk gelinciran kayu kerana saiz aksara berbilang skala dan nisbah bidang, ketumpatan aksara yang tinggi dan jarak aksara dengan aksara yang dekat. Dalam kajian ini, kami mencadangkan rangka kerja pengesanan aksara dan penyetempatan berasaskan U-Net baharu yang mempelajari kawasan watak dan sempadan antara aksara. Kaedah yang dicadangkan meningkatkan prestasi pembelajaran kawasan watak dengan mempelajari sempadan menegak dan mendatar antara aksara secara serentak. Tambahan pula, dengan menambahkan pemprosesan pasca yang mudah dan kos rendah menggunakan kawasan sempadan aksara yang dipelajari, adalah mungkin untuk mengesan lokasi sekumpulan aksara dalam kejiranan yang dekat dengan lebih tepat. Dalam kajian ini, kami membina set data slip kayu. Eksperimen menunjukkan bahawa kaedah yang dicadangkan mengatasi kaedah pengesanan aksara sedia ada, termasuk kaedah pengesanan aksara terkini untuk dokumen sejarah.
Hojun SHIMOYAMA
Kansai University
Soh YOSHIDA
Kansai University
Takao FUJITA
Kansai University
Mitsuji MUNEYASU
Kansai University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Salinan
Hojun SHIMOYAMA, Soh YOSHIDA, Takao FUJITA, Mitsuji MUNEYASU, "U-Net Architecture for Ancient Handwritten Chinese Character Detection in Han Dynasty Wooden Slips" in IEICE TRANSACTIONS on Fundamentals,
vol. E106-A, no. 11, pp. 1406-1415, November 2023, doi: 10.1587/transfun.2023SMP0007.
Abstract: Recent character detectors have been modeled using deep neural networks and have achieved high performance in various tasks, such as text detection in natural scenes and character detection in historical documents. However, existing methods cannot achieve high detection accuracy for wooden slips because of their multi-scale character sizes and aspect ratios, high character density, and close character-to-character distance. In this study, we propose a new U-Net-based character detection and localization framework that learns character regions and boundaries between characters. The proposed method enhances the learning performance of character regions by simultaneously learning the vertical and horizontal boundaries between characters. Furthermore, by adding simple and low-cost post-processing using the learned regions of character boundaries, it is possible to more accurately detect the location of a group of characters in a close neighborhood. In this study, we construct a wooden slip dataset. Experiments demonstrated that the proposed method outperformed existing character detection methods, including state-of-the-art character detection methods for historical documents.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2023SMP0007/_p
Salinan
@ARTICLE{e106-a_11_1406,
author={Hojun SHIMOYAMA, Soh YOSHIDA, Takao FUJITA, Mitsuji MUNEYASU, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={U-Net Architecture for Ancient Handwritten Chinese Character Detection in Han Dynasty Wooden Slips},
year={2023},
volume={E106-A},
number={11},
pages={1406-1415},
abstract={Recent character detectors have been modeled using deep neural networks and have achieved high performance in various tasks, such as text detection in natural scenes and character detection in historical documents. However, existing methods cannot achieve high detection accuracy for wooden slips because of their multi-scale character sizes and aspect ratios, high character density, and close character-to-character distance. In this study, we propose a new U-Net-based character detection and localization framework that learns character regions and boundaries between characters. The proposed method enhances the learning performance of character regions by simultaneously learning the vertical and horizontal boundaries between characters. Furthermore, by adding simple and low-cost post-processing using the learned regions of character boundaries, it is possible to more accurately detect the location of a group of characters in a close neighborhood. In this study, we construct a wooden slip dataset. Experiments demonstrated that the proposed method outperformed existing character detection methods, including state-of-the-art character detection methods for historical documents.},
keywords={},
doi={10.1587/transfun.2023SMP0007},
ISSN={1745-1337},
month={November},}
Salinan
TY - JOUR
TI - U-Net Architecture for Ancient Handwritten Chinese Character Detection in Han Dynasty Wooden Slips
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1406
EP - 1415
AU - Hojun SHIMOYAMA
AU - Soh YOSHIDA
AU - Takao FUJITA
AU - Mitsuji MUNEYASU
PY - 2023
DO - 10.1587/transfun.2023SMP0007
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E106-A
IS - 11
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - November 2023
AB - Recent character detectors have been modeled using deep neural networks and have achieved high performance in various tasks, such as text detection in natural scenes and character detection in historical documents. However, existing methods cannot achieve high detection accuracy for wooden slips because of their multi-scale character sizes and aspect ratios, high character density, and close character-to-character distance. In this study, we propose a new U-Net-based character detection and localization framework that learns character regions and boundaries between characters. The proposed method enhances the learning performance of character regions by simultaneously learning the vertical and horizontal boundaries between characters. Furthermore, by adding simple and low-cost post-processing using the learned regions of character boundaries, it is possible to more accurately detect the location of a group of characters in a close neighborhood. In this study, we construct a wooden slip dataset. Experiments demonstrated that the proposed method outperformed existing character detection methods, including state-of-the-art character detection methods for historical documents.
ER -