IIIT 5K

IIIT 5k字数据集是从Google图片搜索中收获的。使用广告牌，招牌，门牌号，门牌，电影海报等查询词来收集图像。该数据集包含来自场景文本和born-digital图像的5000裁剪单词图像。数据集分为火车和测试部分。此数据集可用于大型词典裁剪单词识别。我们还通过此数据集提供了50万多个词典单词的词典。

IIIT 5k-word Dataset was collected from Google Images Search. Images were gathered using query terms such as billboards, shop signs, house numbers, street nameplates, movie posters and other similar terms. This dataset contains 5,000 cropped word images sourced from both scene text and born-digital images. It is split into training and test subsets. This dataset is applicable for large-vocabulary cropped word recognition tasks. Additionally, a lexicon comprising over 500,000 dictionary words is provided alongside this dataset.

搜集汇总

背景与挑战

背景概述

IIIT 5K数据集包含5000个从Google图片搜索收集的裁剪单词图像，适用于场景文本和born-digital图像的大型词典裁剪单词识别，由海得拉巴国际信息技术研究所在2012年发布。