We will host a Kaggle competition "Kuzushiji Recognition: Opening the Door to A Thousand Years of Japanese Culture" from July 19 to October 14, 2019.
Kaggle competition page: Kuzushiji Recognition - Opening the door to a thousand years of Japanese culture
We will hold a symposium also as the award ceremony of Kaggle competition.
Result of the Competition
Top 5 teams of the Leaderboard will receive the prize. The top 10 teams are as follows.
|1||tascj||0.950||1st place solution|
|2||Konstantin Lopuhin||0.950||2nd place solution overview: detection + full-page classification|
|3||Kenji||0.944||3rd place solution overview: 2-stage + FalsePositive Predictor|
|4||YoudaoOCR||0.942||4th place solution|
|5||See--||0.940||5th place solution overview: One stage CenterNet|
|7||K_mat||0.934||7th place solution|
|8||t-hanya||0.920||8th place solution: Two stage & kuzushiji data augmentation|
|9||Ollie, Nanashi, and Tom||0.910||Top 9th Solution: Simple but complete approach.|
The following summarize the history of kuzushiji recognition, and other basic information.
Examples of Machine Learning (AI) for Kuzushiji
The following is the list of application and services for kuzushiji using machine learning (AI).
Important notice: the following datasets are different from the dataset used in Kaggle competition. Participants of the competition should download the dataset from the Kaggle website.
- Asanobu KITAMOTO, Tarin CLANUWAT, Tomo MIYAZAKI, Kazuaki YAMAMOTO, "Analysis of Character Data: Potential and Impact of Kuzushiji Recognition by Machine Learning", Journal of IEICE (Institute of Electronics, Information, and Communication Engineers), Vol. 102, No. 6, pp. 563-568, doi:10.20676/00000349, June 2019 (in Japanese)
- Asanobu KITAMOTO, "The Development of Data-Driven Humanities Research and Kuzushiji Recognition by AI", Monthly J-LIS, Vol. 6, No. 8, pp. 36-39, doi:10.20676/00000352, 2019-11 (in Japanese)
- Tarin CLANUWAT, Alex LAMB, Asanobu KITAMOTO, "KuroNet: Pre-Modern Japanese Kuzushiji Character Recognition with Deep Learning", 15th International Conference on Document Analysis and Recognition (ICDAR2019), pp. (in press), September 2019
- Tarin CLANUWAT, Alex LAMB, Asanobu KITAMOTO, "End-to-End Pre-Modern Japanese Character (Kuzushiji) Spotting with Deep Learning", Proceedings of IPSJ SIG Computers and the Humanities Symposium 2018, pp. 15-20, December 2018 [ Paper ]
- Tarin CLANUWAT, Mikel BOBER-IRIZAR, Asanobu KITAMOTO, Alex LAMB, Kazuaki YAMAMOTO, David HA, "Deep Learning for Classical Japanese Literature", NeurIPS 2018 Workshop on Machine Learning for Creativity and Design, December 2018 [ Paper ]
- Asanobu KITAMOTO, Tarin CLANUWAT, Alex LAMB, Mikel BOBER-IRIZAR, "Progress and Results of Kaggle Machine Learning Competition for Kuzushiji Recognition", Proceedings of IPSJ SIG Computers and the Humanities Symposium 2019, pp. (in press), December 2019 (in Japanese)
- Anh Duc Le, Tarin CLANUWAT, Asanobu KITAMOTO, "A human-inspired recognition system for pre-modern Japanese historical documents", IEEE Access, pp. 1-7, doi:10.1109/ACCESS.2019.2924449, June 2019
- Asanobu KITAMOTO, Kazuaki YAMAMOTO, "Construction of trans-disciplinary data platform that explores open data in the humanities", Proceedings of IPSJ SIG Computers and the Humanities Symposium 2016, pp. 117-124, 2016-12 (in Japanese) [ Paper ]
- 2nd CODH Seminar: Old Japanese Character Challenge - Future of Machine Recognition and Human Transcription -
- 【プレスリリース】「くずし字」の認識に世界のAI研究者・技術者が挑戦（2019/07/10） (Joint Support-Center for Data Science Research)
- 「くずし字」の認識に世界のAI研究者・技術者が挑戦―全世界的コンペティションをKaggleで7月から開催― (National Institute of Informatics)
第23回 PRMUアルゴリズムコンテスト くずし字認識チャレンジ2019 (2019-08-31締切）