CRHKS- Character Recognition in Handwritten Kadamba Script: A Dataset-Based Approach

Abstract

The Kadamba manuscript is an early South Indian script derived from the Brahmi script

developed in the fourth century by the Kadamba dynasty. It is an ancient and historically significant writing

system from South India, yet its preservation and computational analysis remain challenging due to the absence

of standardized digital representation.It plays a pivotal role in the development of Kannada and Telugu scripts

and is frequently found in early inscriptions.This paper presents a novel dataset specifically designed for the

recognition of handwritten Kadamba script. The Kadamba script data set comprises 29 consonants, 5 vowels,

and 10 numerals. Data were collected from 100 participants representing a diverse range of ages and genders.

Participants were provided with sample templates and instructed to write isolated characters using regular

pens on A4 sheets.To emulate the appearance of traditional manuscripts, characters were written by individuals

of various backgrounds. The data set collected was stored in both CSV and image formats. Each handwritten

sheet was scanned and processed through a structured pipeline to enhance image quality and ensure uniformity.

To preserve the structural integrity of the script, the samples underwent digitization and preprocessing steps,

including adaptive binarization and contour-based segmentation. These processed samples were then used to

build machine learning models. This work addresses the lack of Kadamba script resources by introducing a

benchmark dataset. This contribution systematically bridges the gap by offering a standardized benchmark

Kadamba script for numeral and vowel recognition, which are highly valuable in the context of manuscript

analysis and research like OCR, Historical and Cul