数据堂—20,000张自然场景OCR描述数据

20,000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 20,000 natural scene OCR description samples, covering Asian and European language families with a total of 14 languages. The data collection scenarios include various settings such as store signs, bus stop signs, posters, road signs, etc., and involve multiple shooting angles. All descriptions are written in English, mainly elaborating on information including text arrangement, text content, color and other related details.

搜集汇总

背景与挑战

背景概述

该数据集包含20,000张自然场景图像及其对应的英文描述，涵盖14种语言，场景包括商店招牌、车站标志等多种环境。描述内容主要涉及文本的排列、内容、颜色等信息，图像为.jpg格式，文本为.txt格式。