Researches advanced in Natural Scenes Text Detection Based on Deep Learning

Abstract

The research on text detection and recognition in natural scenes is of great significance for obtaining information from scenes. Thanks to the rapid development of convolutional neural networks and the continuous proposal of scene text detection methods based on deep learning, breakthroughs have been made in the recognition accuracy and speed of scene texts. This paper mainly sorts, analyzes and summarizes the scene text detection method based on deep learning and its development. Firstly, the related research background and significance of scene text detection are discussed. Then, the second part is the elaboration of some main technical research routes of scene text detection. According to the timeline of the detection methods, the specific contents of various text detection models are further introduced. Thirdly, this paper compares and analyzes the experimental results of different models. Furthermore, improvements of some models with relationship, effects, advantages and disadvantages and expectations are further introduced. Finally, the challenges and development trends of scene text detection technology based on deep learning are summarized.