KuroNet: Pre-Modern Japanese Kuzushiji Character Recognition with Deep Learning

@article{Clanuwat2019KuroNetPJ,
  title={KuroNet: Pre-Modern Japanese Kuzushiji Character Recognition with Deep Learning},
  author={Tarin Clanuwat and Alex Lamb and Asanobu Kitamoto},
  journal={2019 International Conference on Document Analysis and Recognition (ICDAR)},
  year={2019},
  pages={607-614}
}
Kuzushiji, a cursive writing style, had been used in Japan for over a thousand years starting from the 8th century. Over 3 millions books on a diverse array of topics, such as literature, science, mathematics and even cooking are preserved. However, following a change to the Japanese writing system in 1900, Kuzushiji has not been included in regular school curricula. Therefore, most Japanese natives nowadays cannot read books written or printed just 150 years ago. Museums and libraries have… 
KuroNet: Regularized Residual U-Nets for End-to-End Kuzushiji Character Recognition
TLDR
The proposed model KuroNet (which builds on Clanuwat et al. in International conference on document analysis and recognition (ICDAR), 2019) outperforms other model for Kuzushiji recognition and achieves higher accuracy while still recognizing entire pages of text using the residual U-Net architecture from adding more regularization.
Predicting the Ordering of Characters in Japanese Historical Documents
TLDR
A few approaches to the task of predicting the sequential ordering of the characters are explored: one using simple hand-crafted rules, another using hand- crafted rules with adaptive thresholds, and another using a deep recurrent sequence model trained with teacher forcing.
Research Challenges in Off-Line Ancient Handwriting Recognition – A Deep Learning Approach
  • Yi Wang, Chen Wang, Bo Chen
  • Computer Science
  • 2020
TLDR
In this paper, some deep learning based approaches are presented to deal with paper-based documents and a framework of such a system is proposed and future directions for the upcoming researchers in the field are proposed.
Restoring and Mining the Records of the Joseon Dynasty via Neural Language Modeling and Machine Translation
TLDR
This work presents a multi-task learning approach to restore and translate historical documents based on a self-attention mechanism, specifically utilizing two Korean historical records, ones of the most voluminous historical records in the world.
KaoKore: A Pre-modern Japanese Art Facial Expression Dataset
TLDR
A new dataset KaoKore is proposed which consists of faces extracted from pre-modern Japanese artwork which demonstrates its value as both a dataset for image classification as well as a creative and artistic dataset, which is explored using generative models.
FAN-MCCD: Fast and Accurate Network for Multi-Scale Chinese Character Detection
TLDR
Focusing on a well-designed neural network architecture and loss function that presents well-classified examples, this work found that detection performance was enhanced for the detection rate, the false positive per character (FPPC), and the F-score in the order of 98.84%, 0.71, and 97.64%, respectively.
Ancient Korean Neural Machine Translation
TLDR
This paper proposes the first ancient Korean neural machine translation model using a Transformer that can improve the efficiency of a translator by quickly providing a draft translation for a number of untranslated ancient documents.
A squeeze and excitation ResNeXt-based deep learning model for Bangla handwritten compound character recognition
TLDR
A deep CNN (Convolutional Neural Network) model using the SE-ResNeXt, which exhibits an average accuracy of 99.82% in recognizing Bangla handwritten compound characters and outperforms the state-of-the-art models by demonstrating higher results.
Recognition of Historical Characters by Combination of Method Detecting Character in Photo Image of Document and Method Separating Block to Characters
TLDR
This paper proposes a method combining a method using deep learning to detect characters in a photo image and a method separating a block into characters, and it is expected that the recognition ratio is improved.
Automatic Evaluation of Difficult-to-Recognize Characters for Efficient Transcription of Japanese Cursive Characters
In this study, we approached Japanese cursive character (kuzushiji) recognition using machine learning to improve the efficiency of their transcription process. In particular, we discussed methods
...
1
2
...

References

SHOWING 1-10 OF 27 REFERENCES
Attempts to recognize anomalously deformed Kana in Japanese historical documents
TLDR
This paper presents methods for three different tasks of recognizing anomalously deformed Kana in Japanese historical documents, which were contested by IEICE PRMU1 2017 and proves the performance of CNN and BLSTM on these tasks.
Deep Learning for Classical Japanese Literature
TLDR
This work introduces Kuz Kushiji-MNIST, a dataset which focuses on Kuzushiji (cursive Japanese), as well as two larger, more challenging datasets, KuzUSHiji-49 and Kuzushaiji-Kanji, which are intended to engage the machine learning community into the world of classical Japanese literature.
One-shot Learning with Memory-Augmented Neural Networks
TLDR
The ability of a memory-augmented neural network to rapidly assimilate new data, and leverage this data to make accurate predictions after only a few samples is demonstrated.
Professor Forcing: A New Algorithm for Training Recurrent Networks
TLDR
The Professor Forcing algorithm, which uses adversarial domain adaptation to encourage the dynamics of the recurrent network to be the same when training the network and when sampling from the network over multiple time steps, is introduced.
Group Normalization
TLDR
Group Normalization can outperform its BN-based counterparts for object detection and segmentation in COCO, and for video classification in Kinetics, showing that GN can effectively replace the powerful BN in a variety of tasks.
Manifold Mixup: Learning Better Representations by Interpolating Hidden States
TLDR
Manifold Mixup achieves large improvements over strong baselines in supervised learning, robustness to single-step adversarial attacks, semi-supervised learning, and Negative Log-Likelihood on held out samples.
A Hybrid RNN Model for Cursive Offline Handwriting Recognition
TLDR
To improve the model performance, the use of specialized Support Vector Machine combined with the original MDRNN in cases of confusion letters to avoid misclassifications is proposed.
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
TLDR
Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin.
Adam: A Method for Stochastic Optimization
TLDR
This work introduces Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments, and provides a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework.
U-Net: Convolutional Networks for Biomedical Image Segmentation
TLDR
It is shown that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks.
...
1
2
3
...