数据堂—4,601张22种票据OCR数据

4,601张22种票据OCR数据采集背景为纯色，数据包括22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 contains 4,601 OCR samples of 22 types of bills, collected under pure-color backgrounds. It covers 22 bill categories and involves multiple provinces, with annotations including line-level quadrilateral labeling and line-level content transcription. This bill OCR dataset can be applied to tasks such as bill recognition and text recognition.

搜集汇总

背景与挑战

背景概述

该数据集包含4,601张22种票据的OCR数据，采集于纯色背景，标注内容包括行级四边形框和文本转录，适用于票据识别和文本识别任务。数据来自多个省份，准确率要求不低于97%。