고문서의 문서 판단, 한자 자형 추출 및 인식 Layout Analysis, Character Detection and Recognition for Historical Document Digitization 古籍中文文字检测、识别及版面分析
- JIN Lianwen(金连文, School of Electronics and Information Engineering, South China University of Technology, China)
We propose an end-to-end trainable framework for restoring historical documents content that follows the correct reading order. In this framework, two branches named character branch and layout branch are added behind an deep learning backbones. The character branch localizes individual characters in a document image and recognizes them simultaneously. Then we adopt a post-processing method togroup them into text lines. The layout branch based on fully convolutional network outputs a binary mask. These two branches can be jointly trained in an end-to-end manner. Furthermore, we propose a re-score mechanism to minimize recognition error. Experiment results on the Chinese historical document TKH-MTH2200 dataset demonstrate the effectiveness of the proposed framework. I will also introduce a real-time Tripitaka OCR system we have built.
我们提出了一个按照正确的阅读顺序识别历史文档的端到端的可训练框架。 在此框架中，在深度学习骨干后面添加了字符分支和布局分支，字符分支在文档图像中定位单个字符并同时进行识别它； 然后，我们采用后处理方法将它们合并为文本行识别结果。布局分支是一个基于全卷积网络的子网络，可输出版面分析结果。 这两个分支可以端到端的方式共同训练。 此外，我们还提出了一种重评分机制以最大程度地减少识别错误。 在中国古籍数据集TKH-MTH2200上的实验结果证明了该方法的有效性。 我还将介绍我们构建的一个实时Tripitaka OCR演示系统。 | 
