---
license: mit
language:
- zh
task_categories:
- image-to-text
pretty_name: CASIA-HWDB2-line
dataset_info:
  features:
  - name: image
    dtype: image
  - name: text
    dtype: string
  splits:
  - name: train
    num_examples: 33401
  - name: validation
    num_examples: 8318
  - name: test
    num_examples: 10441
  dataset_size: 52160
tags:
  - atr
  - htr
  - ocr
  - modern
  - handwritten
---

# CASIA-HWDB2 - line level

## Table of Contents
- [CASIA-HWDB2 - line level](#casia-hwdb2-line-level)
  - [Table of Contents](#table-of-contents)
  - [Dataset Description](#dataset-description)
    - [Dataset Summary](#dataset-summary)
    - [Languages](#languages)
  - [Dataset Structure](#dataset-structure)
    - [Data Instances](#data-instances)
    - [Data Fields](#data-fields)

## Dataset Description

- **Homepage:** [CASIA-HWDB2](http://www.nlpr.ia.ac.cn/databases/handwriting/Download.html)
- **Paper:** [Online and offline handwritten Chinese character recognition: Benchmarking on new databases](https://www.sciencedirect.com/science/article/abs/pii/S0031320312002919)
- **Point of Contact:** [TEKLIA](https://teklia.com)

## Dataset Summary

The offline Chinese handwriting database (CASIA-HWDB2) was built by the National Laboratory of Pattern Recognition (NLPR), Institute of Automation of Chinese Academy of Sciences (CASIA). 
The handwritten samples were produced by 1,020 writers using Anoto pen on papers, such that both online and offline data were obtained.

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

### Languages

All the documents in the dataset are written in Chinese.

## Dataset Structure

### Data Instances

```
{
  'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=1244x128 at 0x1A800E8E190,
  'text': '2007年高校招生录取工作即将陆续展开,教育部有关负责人'
}
```

### 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.