数据堂—71,535张英文自然场景OCR数据

71,535张英文自然场景OCR数据采集场景为英国和美国实地采集。数据多样性包括多种场景、多种拍摄角度、多种光照条件。 在标注方面，对英文OCR数据进行行级&单词级&字符级矩形框（或四边形框）标注，并对文本内容进行转写。本套英文自然场景OCR数据可用英文自然场景下的OCR任务。

71,535 English natural scene OCR samples were collected via field surveys in the UK and the United States. The dataset features high diversity, covering various scenarios, shooting angles, and lighting conditions. For annotation, the English OCR data was annotated with rectangular (or quadrilateral) bounding boxes at the line-level, word-level, and character-level, along with text transcriptions. This English natural scene OCR dataset is applicable to OCR tasks in English natural scenes.

搜集汇总

背景与挑战

背景概述

该数据集包含71,535张英文自然场景OCR图像，适用于英文自然场景的OCR任务。图像采集自英国和美国的不同场景，标注内容包括行级、词级和字符级的边界框及转录文本，具有多样性和高准确性。