Deciphering Ancient Tamil Epigraphy: A Deep Learning Approach for Vatteluttu Script Recognition
Year: 2025
Type: article
Abstract: Vatteluttu (Vattezhuthu) script, prevalent between the 3rd and 8th century CE, played a crucial role in documenting early Chola and Pandya history. However, despite its historical significance, thousands of inscriptions remain untranslated, hindering a comprehensive understanding of early South Indian heritage. This research addresses these challenges by developing a deep learning-based approach for digitizing and recognizing Vatteluttu characters. Using a dataset of 1,800 segmented images representing 28 characters, the study integrates advanced preprocessing techniques and a Siamese CNN-RNN architecture to classify ancient scripts with an overall accuracy of 98%. The findings demonstrate the feasibility of automated transcription for ancient scripts, offering a robust framework for preserving and enhancing research on South Indian heritage.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $550
OpenAlex ID: W4409085889