数据堂—101人4,538张日语手写体OCR数据

101人4,538张日语手写体OCR数据文字载体为A4纸。该数据内容涵盖社会民生、娱乐、旅游、体育赛事、电影、作曲等领域。在标注方面，对文本进行字符级矩形框标注、字符级内容转写和行级矩形框标注、行级内容转写。本套数据可用于日语手写体OCR任务。

A dataset containing 4,538 Japanese handwritten text samples collected from 101 individuals, with all samples written on A4-sized paper. The content of these samples covers various domains including social livelihood, entertainment, tourism, sports events, movies, and music composition. In terms of annotation, both character-level and line-level annotations are provided: character-level bounding box labeling, character-level content transcription, line-level bounding box labeling, and line-level content transcription. This dataset can be utilized for Japanese handwritten OCR tasks.

搜集汇总

背景与挑战

背景概述

该数据集包含4,538张由101人书写的日语手写体图像，格式为.jpg文件，并附有.json格式的标注，涵盖社会民生、娱乐、旅游等多个领域。标注内容包括字符级和行级的矩形框定位及内容转录，适用于日语手写OCR测试任务。