Computer Vision and Deep learning algorithms for the automatic processing of Wasan documents
-
- SUZUKI Toya
- University of Yamagata
-
- DIEZ Yago
- University of Yamagata
-
- VILA Marius
- University of Girona
-
- WAKI Katsushi
- University of Yamagata
Bibliographic Information
- Other Title
-
- コンピュータービジョンとディープラーニングアルゴリズムを利用した和算書の自動処理
Abstract
<p>Wasan documents are a type of mathematical texts unique to Japan dating back to the Edo period. These Wasan documents offer a variety of knowledge and are of great cultural and historical importance. The research I am working on is to build a database of searchable Wasan's figure problems. The purpose of this project is to provide those who are interested in Wasan or who need graphics for educational purposes to meet their requirements. In this paper, we present an algorithm to automatically detect and classify kanji elements in the Wasan documents in order to identify the special kanji "ima" that signals the start of the description of the Wasan’s figure problem. Specifically, the problem of detecting and recognizing manually scanned low-quality Wasan documents and handwritten kanji, including conventional computer vision technology and deep learning for noise reduction, page angle correction, kanji detection and classification. And achieved a 93% success rate. Future research aims to improve the detection accuracy of kanji, and also to improve the classification accuracy of kanji by using different CNN models and data sets.</p>
Journal
-
- Proceedings of the Annual Conference of JSAI
-
Proceedings of the Annual Conference of JSAI JSAI2020 (0), 4Rin110-4Rin110, 2020
The Japanese Society for Artificial Intelligence
- Tweet
Details
-
- CRID
- 1390285300166351616
-
- NII Article ID
- 130007857447
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- CiNii Articles
-
- Abstract License Flag
- Disallowed