An Overview of State-of-the-art Solutions for Scene Text Detection

Abstract

Scene text detection is a task of identifying text regions and labeling them with bounding boxes in a complex background. It has received a lot of attention recently and has become far from unsolvable due to progress of deep learning for computer vision and also due to rapid development of computer hardware which is able to process complex neural networks. Some of the most common challenges that make this task difficult are irregular text shapes, text interferences, very complex background, different text sizes and low image quality. This paper presents an overview of state-of-the-art solutions for scene text detection where ICDAR 2015 was used as a benchmark dataset. We compare solutions with respect to precision, recall and F-score.