BN-HTRd
收藏资源简介:
BN-HTRd数据集是由Premier University和University of Chittagong合作创建,用于离线孟加拉语手写文本识别和行分割。该数据集基于BBC孟加拉语新闻语料库,包含788张手写页面,由约150名不同作者书写,总计108,147个手写单词。创建过程中,首先收集并整理了新闻文本,然后分发给不同背景的作者进行手写。数据集主要用于解决手写文档的自动识别问题,支持从文档到单词级别的识别任务,如端到端文档识别和单词分割等。
The BN-HTRd dataset was collaboratively developed by Premier University and the University of Chittagong for offline Bengali handwritten text recognition and line segmentation. Built upon the BBC Bengali news corpus, this dataset consists of 788 handwritten pages created by approximately 150 diverse authors, totaling 108,147 handwritten words. During its development, news texts were first collected and curated, then distributed to authors with varied backgrounds for handwritten transcription. This dataset is primarily designed to address automatic recognition challenges for handwritten documents, supporting recognition tasks ranging from document-level to word-level, such as end-to-end document recognition and word segmentation.




