EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Automated Classification of Sharda Script Vowels and Numerals Using CNN and Transfer Learning

Preeti Devi

Journal of Intelligent Decision Making and Information Science · 10.10.2026 · 3 (17s) · 2857-2869

DOI: 10.17762/jidmis.v3.5695 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 10.10.2026. Verfasst von Preeti Devi. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Script is the key component of language. The identification of script plays a vital role in language translation, digitization of manuscripts, and text analysis. Various machine learning (ML) and deep learning (DL) models are used for script identification; however, certain challenges persist, including the misclassification of similar-looking characters, complex ligatures, and limited datasets. Sharada is the sacred script of the Kashmir region, dating back to the 8th century CE, and has largely been replaced by Devanagari after the 12th century; however, it remains significant for studying historical texts. Sharada alphabet consists of vowels, consonants, numerals, and modifiers. The research idea is to develop and assess a novel automated handwritten Sharada script recognition system for the precise classification of numerals and vowels numerals into their corresponding classes. The proposed work is divided in two parts: (1) a custom Convolutional Neural Network (CNN) model is trained from scratch for Sharada-numerals-vowels, and (2) two prominent transfer learning architectures InceptionV3 and MobileNetV2 are used and compared their performance against custom CNN. The classification performance of models has been tested on a randomly selected 20% of the dataset, achieving an accuracy of 99.94% (custom CNN), 99.97% (InceptionV3) and 99.92% (MobileNetV2). This research can furthur be carried out for classifying complete Sharda script character set including its consonants and modifiers that will helpful in identification of other ancient scripts.

Zitieren

APA

Devi, P. (2026). Automated Classification of Sharda Script Vowels and Numerals Using CNN and Transfer Learning. Journal of Intelligent Decision Making and Information Science, 3(17s), 2857-2869. https://doi.org/10.17762/jidmis.v3.5695

Vancouver

Devi P. Automated Classification of Sharda Script Vowels and Numerals Using CNN and Transfer Learning. Journal of Intelligent Decision Making and Information Science. 2026;3(17s):2857-2869. doi:10.17762/jidmis.v3.5695

BibTeX
@article{Devi2026,
  author = {Devi, Preeti},
  title = {Automated Classification of Sharda Script Vowels and Numerals Using CNN and Transfer Learning},
  journal = {Journal of Intelligent Decision Making and Information Science},
  year = {2026},
  volume = {3},
  number = {17s},
  pages = {2857--2869},
  doi = {10.17762/jidmis.v3.5695}
}
Kostenlos starten →