Researchers have developed an iterative fine-tuning approach for optical character recognition (OCR) specifically designed for complex historical Sanskrit manuscripts. This method adapts to the unique characteristics of each manuscript, reducing the need for extensive manual annotation. The process has been used to digitize three Sanskrit manuscripts, creating a new dataset with detailed layout and Unicode annotations in PAGE-XML format. The study also benchmarks the performance of leading Multi-modal Large Language Models on this dataset. AI
IMPACT This research could enable more efficient and accurate digitization of historical texts, making them more accessible for scholarly study.
RANK_REASON The item describes a new research paper detailing a novel method for OCR on historical manuscripts. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
