Natural scene text detection algorithm based on the regional proposal

Abstract

Text detection in natural scenes is widely used in various fields. The combination of deep learning models and text detection improves the accuracy and efficiency of detection. Its algorithm is also endless, with continuous optimization and improvement. On this basis, some algorithms for text detection in natural scenes are summarized from the two aspects of the Faster R-CNN method and SSD method based on regional suggestion. These algorithms and experimental data are analyzed and compared to further summarize the challenges faced by text detection in natural scenes and the direction of improvement.