一种用于文本检测和文本识别的单一神经网络

A Single Neural Network for Text Detection and Text Recognition

自然场景中文本的自动检测与读取是一个具有挑战性但尚未完全解决的任务,近几年来,已经出现了文本检测和文本识别的系统。采用了一种用于文本检测和文本识别的单一神经网络STN-OCR,可以进行端到端的优化。STN-OCR是一个集成和共同学习空间变换网络[1]的网络,可以学习检测图像中的文本区域,还可以使用文本识别网络来识别文本区域并识别其文本内容。

In this paper,a single neural network STN-OCR for text detection and text recognition is used for end-to-end optimization.STN-OCR is a network of integrated and common learning spatial transformation networks that can learn to detect text regions in images.Text recognition networks can also be used to identify text regions and identify their text content.