Friday, October 18, 2019

Handwritten Manuscripts into printed, editable text – First kind of Database


Handwritten Manuscripts into printed, editable text – First kind of Database

City gets first-of-its-kind database to decode palm-leaf manuscripts

Hyderabad: Researchers at IIITH (International Institute of Information Technology, Hyderabad) have created the first-ever large-scale database of ancient palm-leaf manuscript images that can identify and label different regions of the document.

They built an algorithm that can read and convert complex, handwritten manuscripts into printed and editable text and trained deep neural networks to automatically identify and isolate different regions of the manuscript images. “Ayurveda and Yoga may have been among the most popular and visible Indian exports to the west, but there exists many more documented shastras or texts in various fields of knowledge, from science and technology to wellness and ecology,” said Ravi Kiran Sarvadevabhatla from the Centre for Visual Information Technology (CVIT), adding that it was the thought of making such a treasure trove more publicly accessible that drove him to embark on a project of cultural and historical importance.

Not only is it difficult to decipher the language on palm-leaf manuscripts, but the lines themselves are so densely packed and tricky to follow, he said.

“In order to have a machine read and convert complex, handwritten manuscripts into printed, editable text, the first step is to let an algorithm identify the layout of the historic images,” said Sarvadevabhatla.

For this purpose, he said they have obtained digitized manuscript images from two sources -- University of Pennsylvania’s Rare Book and Manuscript Library and BHOOMI, an assorted collection of images sourced from multiple oriental research institutes and libraries across India with the help of manual annotations, trained deep neural networks to automatically identify and isolate different regions of the manuscript images.

Source | Business Line | 18th October 2019

Regards

Mr. Pralhad Jadhav  
Research Scholar (IGNOU)
Senior Manager @ Knowledge Repository  
Khaitan & Co 
Twitter Handle | @Pralhad161978
Mobile @ 9665911593

No comments:

Post a Comment