场景文字识别算法的研究

文字识别是一种通用的图像理解技术,对信息检索、车牌识别和自动驾驶等应用的研究有着重要意义。随着神经网络的伟大复兴,场景文字识别任务得到了很大推动,近年来涌现了许多基于深度学习的文字识别算法。本文提出了一种基于特征融合的CRNN改进算法,使用三个通用的文字识别数据集从识别准确率、运行效率和模型大小三个方面进行分析。实验结果表明该算法在提高准确率的同时,运行效率也有所提高。

Researches on Scene Text Recognition Based on Deep Learning

Text recognition is a general technology of image understanding, which is of great significance for the research of information retrieval, license plate recognition, and automatic driving. With the great revival of neural networks, scene text recognition tasks have been greatly promoted. In recent years, many text recognition algorithms based on deep learning have emerged. This paper proposes an improved CRNN algorithm based on feature fusion. It uses three common text recognition datasets to analyze from three aspects: recognition accuracy, operating efficiency and model size. The experiment results show that the proposed algorithm improved the accuracy and efficiency.