A Deep Learning-Based Framework for Marathi OCR and Automated Braille Translation

Madhav A. Kankhar; C Namrata Mahender

doi:10.55524/ijircst.2026.14.4.2

Abstract

This paper presents a deep learning-based framework for converting printed Marathi textbook content into Braille to improve accessibility for visually impaired readers. The proposed system integrates image pre-processing, optical character recognition, character segmentation, convolutional neural network-based classification, Unicode reconstruction, and automated Braille translation into a unified pipeline. Before any character recognition can happen, the scanned pages of Marathi textbooks go through a series of preparation steps. The images are first converted to grayscale, then cleaned up to remove noise, made black and white through Binarization, straightened if they're slightly tilted, and finally sharpened for better clarity. Only after this groundwork is done does the actual character extraction and recognition begin. Once the characters are identified, they're pieced back together into proper Marathi Unicode text. From there, a rule-based system maps each character to its corresponding Braille symbol keeping the process structured and predictable. When tested on a real set of scanned textbook pages, the system performed well, achieving strong character recognition accuracy and producing Braille output that's genuinely reliable for printed educational content. The combination of deep learning with Unicode-aware post-processing turned out to make a meaningful difference in how accurately the Marathi text is read and converted. What makes this work particularly promising is that it's not just accurate it's also scalable. It offers a practical path toward building accessible learning materials, and with some further development, the same approach could be adapted for other Indian languages or applied to documents with more complex layouts down the line.

Keywords

Marathi Braille, CNN, OCR, Text-to-Braille Conversion, Devanagari Script, Machine Learning, Deep Learning

References

Braille, "Braille - Wikipedia, the free encyclopedia." Available from: http://www.Wikipedia.com/Braille
S. Arora, L. Malik, S. Goyal, D. Bhattacharjee, M. Nasipuri, and O. Krejcar, "Devanagari character recognition: A comprehensive literature review," IEEE Access, vol. 13, pp. 1249-1284, 2024. Available from: https://doi.org/10.1109/ACCESS.2024.3520248
N. Gautam, S. S. Chai, and M. Gautam, "The dataset for printed Brahmi word recognition," in Micro-Electronics and Telecommunication Engineering: Proceedings of 3rd ICMETE 2019, Singapore: Springer, 2020, pp. 125-133. Available from: https://doi.org/10.1007/978-981-15-2329-8_13
S. Singh, N. K. Garg, and M. Kumar, "Feature extraction and classification techniques for handwritten Devanagari text recognition: A survey," Multimedia Tools and Applications, vol. 82, no. 1, pp. 747-775, 2023. Available from: https://doi.org/10.1007/s11042-022-13318-9
Acharya, "Introduction to Braille," Multilingual Computing for Literacy and Education. Available from: https://www.acharya.gen.in/
N. B. Jariwala and B. C. Patel, "Multi-Oriented Gujarati Characters Recognition: A Review," International Journal of Research in Computer Science and Information Technology, vol. 2, no. 2(A), pp. 1-5, Mar. 2014. Available from: https://tinyurl.com/8sn5npxx
N. Bhaskar, K. P. Ashritha, P. Tupe-Waghmare, and K. S. Varadaraj, "A Deep Learning-based Text-to-Braille Translation Framework: A Case Study on an Indic Script," in Proc. 6th Int. Conf. Image Processing and Capsule Networks (ICIPCN), Jan. 2026, pp. 771-776. Available from: https://doi.org/10.1109/ICIPCN67432.2026.11439046
N. Khan, A. A. H. Ab Rahman, S. M. Rizvi, I. Y. Alshareef, M. N. Marsono, M. P. Bakht, et al., "Systematic literature review of machine learning models and applications for text recognition," IEEE Access, 2025. Available from: https://doi.org/10.1109/ACCESS.2025.3618109
S. M. Obaidullah, C. Halder, K. C. Santosh, N. Das, and K. Roy, "PHDIndic_11: Page-level handwritten document image dataset of 11 official Indic scripts for script identification," Multimedia Tools and Applications, vol. 77, no. 2, pp. 1643-1678, 2018. Available from: https://doi.org/10.1007/s11042-017-4373-y
S. Thadchanamoorthy, N. D. Kodikara, H. L. Premaretne, U. Pal, and F. Kimura, "Tamil handwritten city name database development and recognition for postal automation," in Proc. 12th Int. Conf. Document Analysis and Recognition (ICDAR), Aug. 2013, pp. 793-797. Available from: https://doi.org/10.1109/ICDAR.2013.162
Nethravathi, C. P. Archana, K. Shashikiran, and R. AG, "Creation of a huge annotated database for Tamil and Kannada OHR," in Proc. Int. Conf. Frontiers in Handwriting Recognition (ICFHR), Nov. 2010, pp. 415-420. Available from: https://doi.org/10.1109/ICFHR.2010.71
U. Bhattacharya and B. B. Chaudhuri, "Databases for research on recognition of handwritten characters of Indian scripts," in Proc. 8th Int. Conf. Document Analysis and Recognition (ICDAR), Aug. 2005, pp. 789-793. Available from: https://doi.org/10.1109/ICDAR.2005.84
Alaei, U. Pal, and P. Nagabhushan, "Dataset and ground truth for handwritten text in four different scripts," International Journal of Pattern Recognition and Artificial Intelligence, vol. 26, no. 4, Art. no. 1253001, 2012. Available from: https://doi.org/10.1142/S0218001412530011
S. M. Obaidullah, C. Halder, N. Das, and K. Roy, "A corpus of word-level offline handwritten numeral images from official Indic scripts," in Proc. 2nd Int. Conf. Computer and Communication Technologies (IC3T), New Delhi, India: Springer, Sep. 2015, pp. 703-711. Available from: https://doi.org/10.1007/978-81-322-2517-1_67
M. Verma, N. Sood, P. P. Roy, and B. Raman, "Script identification in natural scene images: A dataset and texture-feature based performance evaluation," in Proc. Int. Conf. Computer Vision and Image Processing (CVIP), Singapore: Springer, Dec. 2016, pp. 309-319. Available from: https://doi.org/10.1007/978-981-10-2107-7_28
N. Khatun and M. N. Baidya, "Transforming Education in India: NEP 2020 and Inclusiveness," Journal of Research & Innovations in Education, vol. 10, no. 1, pp. 13-26, 2024. Available from: https://publications.cukashmir.ac.in/index.php/jrie/article/view/122
M. A. Kankhar and C. N. Mahender, "A Comprehensive Study of Tactile Education System for Visual Impaired People," in Proc. Int. Conf. Recent Advancements and Modernisations in Sustainable Intelligent Technologies and Applications (RAMSITA), May 2025, pp. 88-99. Available from: https://www.atlantis-press.com/proceedings/ramsita-25/126011501
Ministry of Education, Government of India, "National Education Policy (NEP)." Available from: https://www.education.gov.in/en/nep-new
Balbharti, "Balbharti eBook Library for Marathi Language." Available from: https://ebooks.ebalbharati.in/pdfs/501000541.pdf

Cites this article as

M. A. Kankhar, C. N. Mahender, "A Deep Learning-Based Framework for Marathi OCR and Automated Braille Translation", International Journal of Innovative Research in Computer Science and Technology (IJIRCST), Vol-14, Issue-4, Page No-10-18, 2026. Available from: https://doi.org/10.55524/ijircst.2026.14.4.2

Corresponding Author

Madhav A. Kankhar

Senior Research Fellow, Department of Computer Science & IT, Dr. Babasaheb Ambedkar Marathwada University, Chhatrapati Sambhajinagar, Maharashtra, India

Download Full Paper

Download PDF

No. of Downloads: 15 | No. of Views: 208

A Deep Learning-Based Framework for Marathi OCR and Automated Braille Translation

Citations

Total View 208

Total Download 15