Abstract
Determining the stroke order of a Chinese character image is challenging, because there is no explicit representation for image to sequence learning. This paper investigates the approach in Chinese character generation given just a few image samples of a specific font. Then, keypoint extraction for stroke decomposition and learning to rank method are proposed for obtaining stroke order. Since the same character can appear in multiple fonts, different font of Chinese character has distinct keypoints. Thus, it brings difficulties in acquiring stroke order. Generative Adversarial Networks (GANs) is introduced to generate lots of Chinese character images with different fonts for training and testing the proposed method. The keypoint ranking model based on stroke extraction combining font transfer based on GANs is proposed to complete this task. Compared to other methods, our method can be accomplished without human annotation as initial hints in prediction stage. The experimental results demonstrate the effectiveness of our method that achieved 0.9667 NDCG in average and up to 29.53% samples are higher than 0.98 NDCG.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Cao Z, Qin T, Liu TY, Tsai MF, Li H. Learning to rank: from pairwise approach to listwise approach. In: Proceedings of the 24th international conference on Machine learning. 2007. p. 129–36.
Chang H, Lu J, Yu F, Finkelstein A. Pairedcyclegan: asymmetric style transfer for applying and removing makeup. In Proceedings of the IEEE conference on computer vision and pattern recognition. 2018. p. 40–48.
Chang J, Gu Y. Chinese typography transfer. [Preprint]. 2017. arXiv:1707.04904.
Hayashi H, Abe K, Uchida S. Glyphgan: Style-consistent font generation based on generative adversarial networks. Knowledge Based System. 2019; 186.
Ioffe S, Szegedy C. Batch normalization: acelerating deep network training by reducing internal covariate shift. In Proceedings of the 32nd international conference on maching learning. 2015. 37:448–456. arXiv: 1502.03167
Isola P, Zhu JY, Zhou T, Efros AA. Image-to-image translation with conditional adversarial networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. 2017. p. 1125–1134.
Järvelin K, Kekäläinen J. Cumulated gain-based evaluation of IR techniques. ACM Trans Inf Syst (TOIS). 2002;20(4):422–46.
Krizhevsky A, Sutskever I, Hinton GE. Imagenet classification with deep convolutional neural networks. Commun ACM. 2017;60(6):84–90.
Lin F, Tang X. Off-line handwritten chinese character stroke extraction. Object Recognit Support User Interact Serv Robots. 2002;3:249–52.
Mueller S, Huebel N, Waibel M, D’Andrea R. Robotic calligraphy-learning how to write single strokes of Chinese and Japanese characters. In IEEE/RSJ international conference on intelligent robots and systems. 2013. p. 1734–1739.
Nair V, Hinton GE. Rectified linear units improve restricted boltzmann machines. In ICML. 2010. p.807–814.
Rumelhart DE, Hinton GE, Williams RJ. Learning internal representations by error propagation: technical report. San Diego: California University; 1985.
Sun Y, Qian H, Xu Y. A geometric approach to stroke extraction for the chinese calligraphy robot. In IEEE international conference on robotics and automation (ICRA). 2014. p. 3207–3212.
Tian Y. zi2zi: Master chinese calligraphy with conditional adversarial networks. 2017. https://kaonashi-tyc.github.io/2017/04/06/zi2zi.html.
Wada K. Labelme: image polygonal annotation with Python. 2016. https://github.com/wkentaro/labelme.
Wen C, Chang J, Zhang Y, Chen S, Wang Y, Han M, Tian Q. Handwritten Chinese font generation with collaborative stroke refinement. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 2021. p. 3882–3891.
Xie N, Hachiya H, Sugiyama M. Artist agent: a reinforcement learning approach to automatic stroke generation in oriental ink painting. IEICE Trans Inf Syst. 2013;96(5):1134–44.
Yao F, Shao G, Yi J. Extracting the trajectory of writing brush in Chinese character calligraphy. Eng Appl Artif Intell. 2004;17(6):631–44.
Zeng J, Chen Q, Liu Y, Wang M, Yao Y. Strokegan: Reducing mode collapse in Chinese font generation via stroke encoding. In: proceedings of AAAI. 2021.
Zhang T, Suen CY. A fast parallel algorithm for thinning digital patterns. Commun ACM. 1984;27(3):236–9.
Acknowledgements
This work was supported by Advanced Institute of Manufacturing with High-tech Innovations, Center for Innovative Research on Aging Society (CIRAS), and Ministry of Science and Technology under grant No.109-2218-E-194-009-.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the topical collection “Applications of Cloud Computing, Data Analytics and Building Secure Networks” guest edited by Rajnish Sharma, Pao-Ann Hsiung and Sagar Juneja.
Rights and permissions
About this article
Cite this article
Li, HT., Jiang, MX., Huang, TT. et al. Font Generation and Keypoint Ranking for Stroke Order of Chinese Characters by Deep Neural Networks. SN COMPUT. SCI. 2, 324 (2021). https://doi.org/10.1007/s42979-021-00717-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s42979-021-00717-2