A Novel Joint Character Categorization and Localization Approach for Character-Level Scene Text Recognition

Scene text recognition has become an active research area in pattern recognition in recent years. Currently, the mainstream approach is image-based sequence model. However, such a model usually cannot yield accurate character-level category and location information. To address this deficiency, in this paper, we propose a novel character-level scene text recognition framework for simultaneously categorizing and localizing characters. Moreover, we present an effective joint learning strategy to help the approach to learn from both character-level annotation and word-level annotation. Extensive experiments on five benchmark data sets, including IIIT-5K, SVT, ICDAR03, ICDAR13, and ICDAR15, show promising results. Especially, we confirm that our proposal is more robust to the text length variation and non-language text.