20,011张自然场景OCR描述数据【数据堂】

20,011张自然场景OCR描述数据，包含亚洲语系与欧洲语系，总计14种语言，采集环境包括商店牌匾、站牌、海报、路标等多种场景，含有多种拍摄角度。描述语言为英语，主要描述文字排列方式、文字内容，颜色等信息

This dataset contains 20,011 natural scene OCR images with corresponding English annotations. It covers 14 distinct languages spanning both Asian and European language families. The images are collected from diverse real-world scenarios including store signs, bus stop signs, posters, road signs and other similar contexts, with multiple shooting angles included. All the English annotations primarily describe the text arrangement, text content, color and other relevant details of the OCR elements in the images.

搜集汇总

背景与挑战

背景概述

该数据集包含20,011张自然场景图片及对应的英文描述，覆盖14种亚洲和欧洲语言，采集于商店牌匾、站牌等多种环境。描述主要关注文字排列、内容、颜色和场景，标注准确率不低于97%。