Contact Author

Dr Qingcai Chen
Shenzhen Graduate School,
Harbin Institute of Technology.
Shenzhen, P.R. China 518055
Email: qingcai.chen@hitsz.edu.cn
Tel: +86 755 26033475

Shusen Zhou
Harbin Institute of Technology.
Shenzhen, P.R. China 518055
Email: zhoushusen@gmail.com

Current Version

1.1

Keywords

HIT-OR3C, Online handwriting, Chinese, characters, documents, Character recognition corpus

Description

Example of characters in the dataset

HIT_OR3C is a dataset of handwritten Chinese characters. Both online and offline information is available. The characters have been collected using a handwriting pad and are recorded and labelled automatically via the handwriting document collection software: OR3C Toolkit. The software used to collect the characters is also made available (supplied version is in Chinese).

The dataset is organised in 5 subsets: 4 subsets of characters [Digit (1-10), Letter (11-62), GB1 (63-3817), GB2 (3818-6825)], and 1 subset of documents.

The 4 subsets of characters contain 6,825 classes produced by 122 subjects and 832,650 samples in total. A single file per subject is provided for online data and a single file per subject for offline data (see below for the file format used). The different subsets are defined as index ranges within these files.

The document corpus corresponds to 10 news articles that contain in total 77,168 samples drawn from 2,442 classes and produced by 20 subjects. The document captured data have been post-processed and split into individual characters, the characters resized to 128 x 128 pixels and stored sequentially in a single image and a single vector file, similarly to the first four subsets.

The dataset contains 909,818 images. The total size of the dataset is 15.5 GB (1125 Mb compressed).

There are three file formats, defined by ourselves and introduced in the related documents. The individual character images are 128 x 128 greyscale.

Metadata

For each image, a label is provided. The labels of digits and letters are encoded in ASCII; the labels of Chinese characters are encoded in GB2312 80. The label file is in every folder and named “labels.txt”.