Datasets:

SIR-Lab
/

DohaScript

language:
  - hi
license: other
task_categories:
  - optical-character-recognition
  - image-classification
  - computer-vision
tags:
  - handwriting
  - devanagari
  - hindi
  - ocr
  - document-intelligence
  - computer-vision
pretty_name: Hindi Handwriting Dataset (Sample)
size_categories:
  - n<1K

Hindi Handwriting Dataset (Sample)

Dataset Description

This repository contains a sample subset of 200 handwritten Hindi (Devanagari) document images from DohaScript, a multi-writer handwriting dataset. Each image corresponds to a unique writer and has been collected with the participant's informed consent.

The dataset is intended for research and educational purposes, particularly in handwriting recognition, Optical Character Recognition (OCR), document intelligence, and related computer vision tasks.

Note: This repository contains only 200 samples. The complete DohaScript dataset comprises 531 handwritten pages from 531 unique writers.

Dataset Statistics

Sample Dataset (This Repository)

Attribute	Value
Language	Hindi (Devanagari)
Images	200
Contributors	200 unique writers
Image Format	JPG
Data Modality	Handwritten document images
Metadata Available	Gender, Age, District, Latitude, Longitude
Personally Identifiable Information	None
Consent Obtained	Yes

Full DohaScript Dataset

Attribute	Value
Total Pages	531
Unique Writers	531
Language	Hindi (Devanagari)
Corpus	Six Hindi Dohas (89 words)

Repository Structure

.
├── images/
│   ├── F-18-GAYA.jpg
│   ├── F-21-MADHUBANI.jpg
│   ├── M-18-JAMMU.jpg
│   └── ...
├── metadata.csv
└── README.md

File Naming Convention

Each image follows the format

X-NN-DDDD.jpg

where DDDD is a variable-length district label.

Fields

Gender (`X`)

M → Male
F → Female

Age (`NN`)

Age of the participant in years.

Examples:

18
23
45

District (`DDDD`)

The participant's hometown district.

Example filenames:

F-18-GAYA.jpg
F-21-MADHUBANI.jpg
M-18-JAMMU.jpg

Metadata

Each handwriting image has a corresponding entry in metadata.csv.

Column	Description
`S. No.`	Unique sample identifier
`Filename`	Image filename
`Gender`	Writer gender (`M` or `F`)
`Age`	Age of the writer (years)
`City`	Home district/city of the participant
`Latitude`	Latitude of the district/city
`Longitude`	Longitude of the district/city

The metadata enables demographic and geographic analyses while preserving participant privacy. No personally identifiable information is included.

Data Collection

One handwritten page per participant
Individual volunteers
Informed consent obtained before data collection
Collected solely for research and educational purposes
Personally identifiable information removed prior to release

Intended Uses

This dataset can be used for research in:

Hindi Handwriting Recognition
Optical Character Recognition (OCR)
Document Intelligence
Writer Identification and Analysis
Computer Vision
Pattern Recognition
Machine Learning
Educational Projects

Out-of-Scope Uses

This dataset must not be used for:

Biometric identification
Surveillance
Profiling individuals
Commercial deployment without permission

Privacy and Ethics

The dataset has been curated following ethical research practices.

No names or personal identifiers are included.
No addresses are provided.
Demographic information is limited to age and gender.
Geographic information is provided only at the district/city level.
All participants voluntarily contributed their handwriting with informed consent.

Users of this dataset are expected to uphold the same ethical standards.

Citation

If you use this dataset in a publication, thesis, benchmark, report, or research project, please cite the DohaScript paper.

Paper

DohaScript: A Large-Scale Multi-Writer Dataset for Continuous Handwritten Hindi Text

arXiv: https://arxiv.org/abs/2602.18089

BibTeX

@article{singh2026dohascript,
  title={DohaScript: A Large-Scale Multi-Writer Dataset for Continuous Handwritten Hindi Text},
  author={Singh, Kunwar Arpit and Prakash, Ankush and Lone, Haroon R},
  journal={arXiv preprint arXiv:2602.18089},
  year={2026}
}

License

This dataset is released for research and educational purposes only.

Commercial use requires explicit permission from the authors.

Redistribution should retain this README and the citation information.

Contact

For questions, permission requests, or issues regarding the dataset, please open an issue in this repository.