10,619张30种票据OCR数据

10,619张30种票据OCR数据采集背景为多种不同背景，数据包括30种票据、多个省份。在标注方面，标注行级四边形标注、行级内容转写。本套票据OCR数据可用于票据识别、文字识别等任务。

This dataset comprises 10,619 document images for 30 types of financial documents, with corresponding OCR data. The data was collected under diverse backgrounds, covers 30 categories of financial documents, and is sourced from multiple provinces. Regarding annotation, it provides line-level quadrilateral bounding box annotations and line-level content transcriptions. This financial document OCR dataset can be applied to tasks including document recognition and text recognition.

搜集汇总

背景与挑战

背景概述

该数据集包含10,619张图像，覆盖30种票据类型，采集自多种背景和多个省份，具有较高的多样性。标注方面提供了行级四边形标注和内容转写，检测框和文本转写精度均不低于97%，适用于票据识别和文字识别等OCR任务。