Abstract
The Devanagari script is one of the most widely used scripts worldwide. The existing deep learning-based optical character recognition system for printed Devanagari scripts using Convolutional Neural Network – Recurrent Neural Network, or CRNN is not robust enough to recognize any randomly printed Devanagari scanned document. At present, the hyper-parameters of the CRNN system are selected randomly either with the trial-and-error or grid search methods. Moreover, there is no optimized way to choose the hyper-parameters of the CRNN, which improves the recognition accuracy for Devanagari documents. Furthermore, the lack of standard Devanagari script datasets has hampered the development of word recognizers. In this paper, the hyper-parameter of the CRNN system is optimized using Taguchi's method of optimization. The performance of the hyper-parameters optimized CRNN system is compared with the current state-of-the-art text recognition CRNN network. The results reveal that the CRNN optimized with Taguchi's method performs better than the CRNN-based systems.
- [1] . The World's 5 Most Commonly Used Writing Systems. https://www.britannica.com/list/the-worlds-5-most-commonly-used-writing-systems.Google Scholar
- [2] . 1999. Partitioning and searching dictionary for correction of optically read Devanagari character strings. In Proceedings of the Fifth International Conference on Document Analysis and Recognition. ICDAR. 653–656.Google Scholar
Cross Ref
- [3] . 2000. Integrating knowledge sources in Devanagari text recognition system. IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans 30, 4 (2000), 500–505.Google Scholar
Digital Library
- [4] . 2015. A segmentation-free approach for printed Devanagari script recognition. In 13th International Conference on Document Analysis and Recognition (ICDAR). IEEE, 946–950.Google Scholar
Digital Library
- [5] . 2017. Printed text recognition using BLSTM and MDLSTM for Indian languages. In Fourth International Conference on Image Information Processing (ICIIP). 1–6.Google Scholar
Cross Ref
- [6] . 2012. Recognition of printed Devanagari text using BLSTM neural network. In Proceedings of the 21st International Conference on Pattern Recognition (ICPR’12). 322–325.Google Scholar
- [7] . 2017. SegNet: A deep convolutional encoder-decoder architecture for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence 39, 12 (2017), 2481–2495.Google Scholar
- [8] . 2006. Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks. In Proceedings of the 23rd International Conference on Machine Learning. 369–376.Google Scholar
Digital Library
- [9] . 2018. Connectionist temporal classification with maximum entropy regularization. Advances in Neural Information Processing Systems (2018), 31.Google Scholar
- [10] . 2017. Benchmarking scene text recognition in Devanagari, Telugu, and Malayalam. In 14th IAPR International Conference on Document Analysis and Recognition (ICDAR). IEEE, 42–46.Google Scholar
Cross Ref
- [11] . 2016. Optimized structure of the traffic flow forecasting model with a deep learning approach. IEEE Trans. Neural Netw. Learn. Syst. 28, 10 (2016), 2371–2381.Google Scholar
Cross Ref
- [12] . 1995. A feature-based scheme for the machine recognition of printed Devanagari script. In Pattern Recognition, Image Processing, and Computer Vision (1995), 163–168.Google Scholar
- [13] . 1997. Printed devanagari script OCR system. Vivek-Bombay-. 10 (1997), 12–24.Google Scholar
- [14] . 2004. Indian script character recognition: A survey. Pattern Recognition 37 (2004), 1887–1899.Google Scholar
Cross Ref
- [15] . 2008. Online handwritten Devanagari word recognition: A holistic approach based on directional chain code feature and HMM. In International Conference on Information Technology, (ICIT'08). IEEE, 23–208.Google Scholar
- [16] . 2001. A complete OCR for printed Hindi text in Devanagari script. Proceedings of Sixth International Conference on Document Analysis and Recognition. IEEE Computer Society.Google Scholar
Cross Ref
- [17] . 2002. Segmentation of touching characters in printed Devanagari and Bangla scripts using fuzzy multi factorial analysis. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 32, 4 (2001), 449–459.Google Scholar
Digital Library
- [18] . 2016. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 770–778.Google Scholar
Cross Ref
- [19] . 2017. Inception-v4, inception-ResNet and the impact of residual connections on learning. In Proceedings of the AAAI Conference on Artificial Intelligence.Google Scholar
Cross Ref
- [20] . 2012. ImageNet classification with deep convolutional neural networks. In Advances in Neural Information Processing Systems (NIPS). 1097–1105.Google Scholar
Digital Library
- [21] . 2021. Multi-task learning for pre-processing of printed Devanagari document images with hyper-parameter optimization of the deep architecture using Taguchi method. Sadhana 46, 3 (2021), 1–24.Google Scholar
Cross Ref
- [22] . 2020. Fast multi-language LSTM-based online handwriting recognition. International Journal on Document Analysis and Recognition (IJDAR) 23, 2 (2020), 89–102.Google Scholar
Digital Library
- [23] . 2019. Collaborative deep neural network for printed text recognition of Indian languages. In Fifth International Conference on Image Information Processing (ICIIP). 252–256.Google Scholar
Cross Ref
- [24] . 2019. Deep neural network hyper-parameter optimization with orthogonal array tuning. In Proceedings of the International Conference on Neural Information Processing. Cham: Springer, 287–295.Google Scholar
Cross Ref
- [25] . 2015. Optimizing deep learning hyper-parameters through an evolutionary algorithm. In Proceedings of the Workshop on Machine Learning in High-Performance Computing Environments. 1–5.Google Scholar
Digital Library
- [26] . 2017. Efficient hyperparameter optimization for deep learning algorithms using deterministic RBF surrogates. In Proceedings of the Thirty-first AAAI Conference on Artificial Intelligence.Google Scholar
Digital Library
- [27] . 2017. Hyperband: A novel bandit-based Sådhanå (2021) approach to hyperparameter optimization. J. Mach. Learn. Res. 18, 1 (2017), 6765–6816.Google Scholar
Digital Library
- [28] . 2012. Random search for hyper-parameter optimization. J. Mach. Learn. Res. 13 (2012), 281–305.Google Scholar
Digital Library
- [29] . 1989. Quality engineering through design optimization. In Quality Control, Robust Design, and the Taguchi Method. Boston, MA: Springer, 77–96.Google Scholar
- [30] . 1993. Taguchi Methods: Design of Experiments. Taguchi Methods Series 4.Google Scholar
- [31] . 1993. Taguchi Methods Explained: Practical Steps to Robust Design. Hoboken: Prentice-Hall.Google Scholar
- [32] . 2016. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence 39, 11 (2016), 2298–2304.Google Scholar
Digital Library
- [33] . 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International Conference on Machine Learning. PMLR, 448–456.Google Scholar
- [34] 2016. You only look once: Unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Google Scholar
- [35] . 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.Google Scholar
- [36] . 2018. Activation functions: Comparison of trends in practice and research for deep learning. arXiv preprint arXiv:1811.03378.Google Scholar
- [37] . 2014. Dropout: A simple way to prevent neural networks from overfitting. The Journal of Machine Learning Research 15, 1 (2014), 1929–1958.Google Scholar
Digital Library
- [38] . 1990. Backpropagation through time: What it does and how to do it. Proceedings of the IEEE 78, 10 (1990), 1550–1560.Google Scholar
Cross Ref
- [39] . 2019. Deep neural network hyperparameter optimization with orthogonal array tuning. In , (Eds). 2019. Neural Information Processing. ICONIP 2019. Communications in Computer and Information Science, vol. 1142. Springer, Cham, Google Scholar
Cross Ref
- [40] . 2021. Hyperparameter optimization of deep learning networks for classification of breast histopathology images. Sensors and Materials 33, 1 (2021), 315–325.Google Scholar
Cross Ref
- [41] . 2020. A statistical design of experiments approach to machine learning model selection in engineering applications. ASME. J. Comput. Inf. Sci. Eng. February 21, 1 (2020), 011008. Google Scholar
Cross Ref
- [42] . 2021. Transfer learning for scene text recognition in Indian languages. In International Conference on Document Analysis and Recognition, Springer, Cham, 182–197.Google Scholar
Digital Library
Index Terms
Hyper Parameter Optimization of CRNN for Printed Devanagari Script Recognition using Taguchi's Method
Recommendations
Touching character segmentation of Devanagari script
ICCCNT '16: Proceedings of the 7th International Conference on Computing Communication and Networking TechnologiesSegmentation of characters is one of the major step in OCR system. Devanagari script is a two dimensional form of symbol. It is very inconvenient to treat each form of character as a separate symbol because such combinations are very large in number. ...
Gujarati Script Recognition
AbstractCharacter recognition is the extraction of printed or handwritten text from images into machine-readable format. The extracted text can be easily edited, modified and efficiently stored. While there are several Optical Character Recognition (OCR) ...
Word and character segmentation in ancient handwritten documents in Devanagari and Maithili scripts using horizontal zoning
AbstractThe word and character segmentation in ancient handwritten Devanagari (Indian) and Maithili (Indian) documents in order to develop any optical character recognition (OCR) systems for the ancient documents are still considered as ...
Highlights- Proposing a novel method of text segmentation in ancient handwritten documents.






Comments