MODI-HChar: Historical MODI Script Handwritten Character Dataset

MODI script was used to write Indian languages as Marathi, Hindi, and Gujarati etc. from 12th century. From 17th century to mid of 19th century MODI was used as administrative script in Maharashtra state (India). Now a days, MODI script users are diminishing away, and countable persons can understand the MODI script. The archaic historical MODI handwritten documents contained important and rare cultural, historic, and administrative type of information which is usable in current era. In the research to train and test the Machine learning system a standard invariant character dataset is required. It is desirable in the development of the character recognition system that proposed approach has generalization proficiencies. The system gives good results if it is trained and tested using a standard invariant dataset. Here a standard invariant dataset of handwritten MODI characters is uploaded. MODI-HChar dataset contains total 57 handwritten MODI character classes images which comprises 10 numerals (0-9), 12 vowels (A – Ah) and 35 consonants (K - Dyn). This dataset includes total 575920 MODI character images as 101100 MODI digit images, 121320 MODI vowel images and 353500 MODI consonant images. This dataset is archived in a zip file. MODI-HChar dataset consists of three main folders as digits, vowels and consonants. Digits folder contains the subfolder for each digit zero to nine. Each of these folders includes 10110 images of the associated MODI digit. Equally vowel folder contains 12 subfolders and consonants folder contains 35 subfolders. And each of these subfolders contains 10110 images of the associated MODI character. The MODI character size is of 170x170 pixels and of 96 dpi. All the images are gray level and having type of the image is JPG. The users of the MODI-HHDoc Dataset must agree that: • Use of the data set is restricted to research purpose only. • No redistribution of the dataset is allowed. • Dataset can be partitioned into training and testing as per the requirement. • In any resultant publications of research that uses the dataset, due credits will be provided to the following publication: - Deshmukh, M. S., Patil, M. P., & Kolhe, S. R. (2015, August). Off-line Handwritten Modi Numerals Recognition using Chain Code. In Proceedings of the Third International Symposium on Women in Computing and Informatics (pp. 388-393).

自12世纪起，莫迪文（MODI script）便被用于书写印度诸语言，涵盖马拉地语、印地语、古吉拉特语等。17世纪至19世纪中叶，莫迪文曾作为印度马哈拉施特拉邦的官方书写体系。时至今日，莫迪文的使用者日渐式微，能够识读该文字的人群已寥寥无几。
古老的莫迪文手写文档承载着珍贵且稀缺的文化、历史与行政类信息，在当代仍具备重要利用价值。在机器学习系统的训练与测试研究中，标准化的不变性字符数据集是不可或缺的基础资源。对于字符识别系统的研发而言，所采用的方法需具备泛化能力；而借助标准化不变数据集开展训练与测试，可使系统获得优异的识别效果。
本次上传的即为标准化手写莫迪文字符不变性数据集——MODI-HChar数据集。该数据集共包含57类手写莫迪文字符图像，其中包括10个数字（0-9）、12个元音（A至Ah）以及35个辅音（K至Dyn）。数据集总计包含575920张莫迪文字符图像，具体为：莫迪数字图像101100张、莫迪元音图像121320张、莫迪辅音图像353500张。本数据集以ZIP压缩包形式归档。
MODI-HChar数据集包含三个主文件夹，分别对应数字、元音与辅音类别。数字文件夹下设有0至9共10个子文件夹，每个子文件夹内含10110张对应莫迪数字的图像。同理，元音文件夹下设12个子文件夹，辅音文件夹下设35个子文件夹，每个子文件夹均包含10110张对应莫迪字符的图像。
莫迪文字符图像的统一规格为：尺寸170×170像素，分辨率96dpi，所有图像均为灰度图，格式为JPG。
使用MODI-HHDoc数据集的用户需遵守以下条款：
• 本数据集仅可用于科研目的。
• 严禁对本数据集进行二次分发。
• 用户可根据研究需求将数据集划分为训练集与测试集。
• 若基于本数据集开展研究并发表相关成果，需对如下文献予以引用致谢：
- Deshmukh, M. S., Patil, M. P., & Kolhe, S. R. (2015, August). Off-line Handwritten Modi Numerals Recognition using Chain Code. In Proceedings of the Third International Symposium on Women in Computing and Informatics (pp. 388-393).