An end-to-end trainable Thai OCR system using deep recurrent neural network รัฐศาสตร์ เฮงประเสริฐ* และ สุรเดช อิณทกรณ์

Main Article Content

รัฐศาสตร์ เฮงประเสริฐ


In this paper, we present an end-to-end trainable model to recognize a Thai word from an image. Compared with other previous Thai OCR system, our system has distinctive features that can handle arbitrary length of Thai word without character segmentation and high-level visual features are learned from data. The neural network model is composed of 2 main modules: Convolutional Layer and Recurrent Neural Network (LSTM).

Article Details



Tanprasert, Chularat, and Thaweesak Koanantakool. 2539. "Thai OCR: a neural network application". Proceedings of Digital Processing Applications (TENCON'96). Vol. 1. IEEE

Shi Baoguang, Xiang Bai, and Cong Yao. 2559. "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): 2298-2304.

Harald Scheidi. 2562. Handwritten Text Recognition (HTR) system implemented with TensorFlow. Handwritten Text Recognition (HTR) system implemented with TensorFlow. แหล่งที่มา : 18 September 2019.

K. Simonyan and A. Zisserman. 2557. “Very deep convolutional networks for large- scale recognition,” arXiv preprint arXiv:1409.1556

Tanprasert, Chularat, et al. 2542. "Improved Mixed Thai & English OCR using Two-step neural net classification." NECTEC technical journal 1 : 41-46.

Sutskever, Ilya, et al. 2556. "On the importance of initialization and momentum in deep learning." International conference on machine learning

Glorot, Xavier, and Yoshua Bengio. 2553. "Understanding the difficulty of training deep feedforward neural networks." Proceedings of the thirteenth international conference on artificial intelligence and statistics.

Lâm, Hoàng. 2562. Line-level Handwritten Text Recognition with TensorFlow. Line-level Handwritten Text Recognition with TensorFlow. source : 28 November 2019

Navarro, Gonzalo. 2554. "A guided tour to approximate string matching." ACM computing surveys (CSUR) 33 (1) : 31-88.