darknight054/indic-mozhi-ocr

Mozhi (Printed Word Images) - Indic OCR Dataset

This folder contains the word-level printed OCR dataset downloaded from the CVIT USODI project page for "Towards Deployable OCR Models for Indic Languages". The data is organized by language and split (train/val/test) and is intended for upload to Hugging Face.

Source

Source page: https://cvit.iiit.ac.in/usodi/tdocrmil.php

Paper: Towards Deployable OCR Models for Indic Languages Authors: Minesh Mathew, Ajoy Mondal, C V Jawahar Conference: International Conference on Pattern Recognition (ICPR)

Languages

The dataset provides word images for 13 languages:

• Assamese
• Bengali
• Gujarati
• Hindi
• Kannada
• Malayalam
• Manipuri
• Marathi
• Oriya (Odia)
• Punjabi
• Tamil
• Telugu
• Urdu

Structure

raw_data_2/
├── assamese/
│   ├── train/
│   │   ├── images/
│   │   ├── train_gt.txt
│   │   └── vocabulary.txt
│   ├── val/
│   │   ├── images/
│   │   └── val_gt.txt
│   └── test/
│       ├── images/
│       └── test_gt.txt
├── bengali/
├── gujarati/
├── hindi/
├── kannada/
├── malayalam/
├── manipuri/
├── marathi/
├── oriya/
├── punjabi/
├── tamil/
├── telugu/
└── urdu/

Note: vocabulary.txt may be present in the train split for some languages.

Transcriptions (*_gt.txt)

Each line in train_gt.txt, val_gt.txt, and test_gt.txt is tab-separated:

images/[image_file].jpeg    text

• images/[image_file].jpeg is the path to the word image relative to the split directory.
• text is the Unicode transcription for that image.

Citation

If you use this dataset in your research, please cite:

@inproceedings{mathew2025towards,
  title={Towards Deployable OCR Models for Indic Languages},
  author={Mathew, Minesh and Mondal, Ajoy and Jawahar, CV},
  booktitle={International Conference on Pattern Recognition},
  pages={167--182},
  year={2025},
  organization={Springer}
}

Acknowledgements

This work is supported by MeitY, Government of India, through the NLTMBhashini project.

License

Please refer to the source page for licensing information.