What does this research mean for the field?

The EpigraphiX-AI system, utilizing neural OCR and topological binarization, achieves 97.4% word accuracy and 99.4% authentication accuracy in digitizing severely degraded historical South Indian palm-leaf manuscripts. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To develop an end-to-end neural optical character recognition pipeline capable of authenticating, cleaning, and transcribing physically degraded historical South Indian palm-leaf manuscripts.

August 27, 2026Open Access

EpigraphiX-AI: Neural Epigraphical OCR, Topological Binarization, Strict Manuscript Authentication, and Multi-Model Intelligence for Degraded Historical Palm-Leaf Manuscripts

View Full Paper

Ask AI

Bookmark

Authors

ASAdarsh S

Discussion

Member takes

Overview

Computational framework evaluation demonstrates high-accuracy text recognition in degraded historical palm-leaf manuscripts, highlighting scalable digital preservation for ancient epigraphical...

Key Points

To develop an end-to-end neural optical character recognition pipeline capable of authenticating, cleaning, and transcribing physically degraded historical South Indian palm-leaf manuscripts.
Designed a system featuring multi-gamut HSV color authentication, fiber-aware neural inpainting via directional Gabor filtering and Photometric Stereo, and O(1) integral Sauvola binarization.
Applied persistent homology topological filtering across Betti numbers to retain character loops, paired with a five-model comparative decision space (SVM, Random Forest, Gaussian Naive Bayes, k-NN, CNN).
Benchmarked the end-to-end pipeline on an archival corpus of 1,250 historical palm-leaf folios.
Achieved a Word Accuracy Rate of 97.4%, Character Accuracy of 98.6%, and a Character Error Rate of 1.4% across the 1,250 folio corpus.
Attained 98.6% precision, 98.4% recall, 99.4% palm-leaf authentication accuracy, and sub-3.8ms per-folio adaptive thresholding latency.

Cite This Study

Adarsh S (2026) studied this question.

synapsesocial.com/papers/6a8fe9da10c91c1e92621dd6 https://doi.org/10.5281/zenodo.22088399

View Full Paper

Ask AI

Bookmark

August 27, 2026Open Access

EpigraphiX-AI: Neural Epigraphical OCR, Topological Binarization, Strict Manuscript Authentication, and Multi-Model Intelligence for Degraded Historical Palm-Leaf Manuscripts

View Full Paper

Ask AI

Bookmark

Authors

ASAdarsh S

Discussion

Member takes

Overview

Computational framework evaluation demonstrates high-accuracy text recognition in degraded historical palm-leaf manuscripts, highlighting scalable digital preservation for ancient epigraphical...

Key Points

To develop an end-to-end neural optical character recognition pipeline capable of authenticating, cleaning, and transcribing physically degraded historical South Indian palm-leaf manuscripts.
Designed a system featuring multi-gamut HSV color authentication, fiber-aware neural inpainting via directional Gabor filtering and Photometric Stereo, and O(1) integral Sauvola binarization.
Applied persistent homology topological filtering across Betti numbers to retain character loops, paired with a five-model comparative decision space (SVM, Random Forest, Gaussian Naive Bayes, k-NN, CNN).
Benchmarked the end-to-end pipeline on an archival corpus of 1,250 historical palm-leaf folios.
Achieved a Word Accuracy Rate of 97.4%, Character Accuracy of 98.6%, and a Character Error Rate of 1.4% across the 1,250 folio corpus.
Attained 98.6% precision, 98.4% recall, 99.4% palm-leaf authentication accuracy, and sub-3.8ms per-folio adaptive thresholding latency.

Cite This Study

Adarsh S (2026) studied this question.

synapsesocial.com/papers/6a8fe9da10c91c1e92621dd6 https://doi.org/10.5281/zenodo.22088399

View Full Paper

Ask AI

Bookmark