johnlockejrr/ICDAR2026-ret-meta-NFD

该数据集是一个多语言图像到文本数据集，包含图像、文本、来源、文件名和类型等特征。数据集支持多种语言，包括拉丁语、法语、意大利语、西班牙语等，主要用于图像到文本的任务。数据集包含一个训练集，大小为35238980155.59字节，包含454457个示例。数据集的来源包括stanford-oval/churro-dataset和zenodo.org的记录。数据集的结构包括多个子目录，每个子目录对应不同的语言和文本类型。

This dataset is a multilingual image-to-text dataset containing features such as images, text, source, filename, and type. The dataset supports multiple languages, including Latin, French, Italian, Spanish, etc., and is primarily used for image-to-text tasks. The dataset includes a training set with a size of 35238980155.59 bytes and 454457 examples. The sources of the dataset include stanford-oval/churro-dataset and records from zenodo.org. The dataset structure includes multiple subdirectories, each corresponding to different languages and text types.