Reading Scene Text in Deep Convolutional Sequences
Abstract
We develop a Deep-Text Recurrent Network (DTRN) that regards scene text reading as a sequence labelling problem. We leverage recent advances of deep convolutional neural networks to generate an ordered high-level sequence from a whole word image, avoiding the difficult character segmentation problem. Then a deep recurrent model, building on long short-term memory (LSTM), is developed to robustly recognize the generated CNN sequences, departing from most existing approaches recognising each character independently. Our model has a number of appealing properties in comparison to existing scene text recognition methods: (i) It can recognise highly ambiguous words by leveraging meaningful context information, allowing it to work reliably without either pre- or post-processing; (ii) the deep CNN feature is robust to various image distortions; (iii) it retains the explicit order information in word image, which is essential to discriminate word strings; (iv) the model does not depend on pre-defined dictionary, and it can process unknown words and arbitrary strings. Codes for the DTRN will be available.
Cite
@article{arxiv.1506.04395,
title = {Reading Scene Text in Deep Convolutional Sequences},
author = {Pan He and Weilin Huang and Yu Qiao and Chen Change Loy and Xiaoou Tang},
journal= {arXiv preprint arXiv:1506.04395},
year = {2015}
}
Comments
To appear in the 13th AAAI Conference on Artificial Intelligence (AAAI-16), 2016