Teklia/NorHand-v3-line

---
license: mit
language:
- nb
task_categories:
- image-to-text
pretty_name: NorHand-v3-line
dataset_info:
features:
- name: image
dtype: image
- name: text
dtype: string
splits:
- name: train
num_examples: 222381
- name: validation
num_examples: 22679
- name: test
num_examples: 1562
dataset_size: 246622
tags:
- atr
- htr
- ocr
- historical
- handwritten
---

# NorHand v3 - line level

## Table of Contents
- [NorHand v3 - line level](#norhand-v3-line-level)
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
- [Languages](#languages)
- [Dataset Structure](#dataset-structure)
- [Data Instances](#data-instances)
- [Data Fields](#data-fields)

## Dataset Description

- **Homepage:** [Hugin-Munin project](https://hugin-munin-project.github.io/)
- **Source:** [Zenodo](https://zenodo.org/records/10255840)
- **Point of Contact:** [TEKLIA](https://teklia.com)

## Dataset Summary

The NorHand v3 dataset comprises Norwegian letter and diary line images and text from 19th and early 20th century.

Note that all images are resized to a fixed height of 128 pixels.

### Languages

All the documents in the dataset are written in Norwegian Bokmål.

## Dataset Structure

### Data Instances

```
{
'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=4300x128 at 0x1A800E8E190,
'text': 'Til Bestyrelsen af'
}
```

### Data Fields

- `image`: a PIL.Image.Image object containing the image. Note that when accessing the image column (using dataset[0]["image"]), the image file is automatically decoded. Decoding of a large number of image files might take a significant amount of time. Thus it is important to first query the sample index before the "image" column, i.e. dataset[0]["image"] should always be preferred over dataset["image"][0].
- `text`: the label transcription of the image.

原始信息汇总

数据集概述

数据集名称

NorHand-v3-line

数据集简介

NorHand v3 数据集包含19世纪和20世纪初的挪威信件和日记行图像及其文本。所有图像都被调整为固定高度128像素。

语言

数据集中的所有文档均以挪威博克马尔语（Norwegian Bokmål）书写。

数据集结构

数据实例

json { image: <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=4300x128 at 0x1A800E8E190>, text: Til Bestyrelsen af }

数据字段

• image: 包含图像的PIL.Image.Image对象。注意，访问图像列时（使用dataset[0]["image"]），图像文件会自动解码。解码大量图像文件可能需要较长时间，因此建议先查询样本索引再访问"image"列，即dataset[0]["image"]应始终优于dataset["image"][0]。
• text: 图像的标签转录文本。

数据集分割

• train: 222381个样本
• validation: 22679个样本
• test: 1562个样本

数据集大小

246622个样本

标签

• atr
• htr
• ocr
• historical
• handwritten