Summary of Scene Text Detection Based on Deep Learning

Abstract

Scene text detection is a general text detection technology, which has become a hot research direction in the field of computer vision and document analysis in recent years, and is widely used in geographic positioning, license plate recognition, unmanned driving, and other fields. Compared with traditional document text detection, scene text changes more dramatically in font, scale, arrangement, and background. Therefore, deep learning technology has become the mainstream method in this field because of its excellent performance, which is helpful to improve the ability of text detection. This paper introduces the main research techniques of natural scene text detection and summarizes the structural characteristics of some classical network models. This paper sorts out, analyzes, and summarizes the running mechanism and performance of various network models for natural scene text detection deep learning-based in recent years. The common public datasets and their application characteristics are listed. Finally, the problems and challenges in scene text detection based on deep learning are discussed and provide an outlook on the future research directions in this field.