Ibteda Digital Library digitizes 1,800 Urdu books using budget cameras and custom AI

A Decade of Digital Archiving

The Ibteda Digital Library, a volunteer initiative by three Pakistani friends, has completed the digitization of 1,800 rare and out-of-print Urdu books. This decade-long project involved using budget Nikon D5300 and D3300 cameras, which collectively accumulated 902,000 shutter counts. The team funded the project and purchased books out of their own pockets.

Overcoming Technical Challenges

The digitization process was initially manual, requiring extensive Photoshop post-processing to achieve archival quality. Challenges included maintaining consistent margins, text size, and perspective correction across diverse books, some of which were centuries-old lithographs. The team eventually developed a machine learning process to automate the post-processing of 526,000 dual-page scans, addressing the complexities of varied book conditions and the unique characteristics of Urdu script.

The Specifics of Urdu Script

A significant hurdle was the nature of Urdu script, primarily Nastaliq, which features flowing characters, numerous dots, diacritics, and small symbols. This made it difficult to distinguish actual writing from photography noise, dirt, or blemishes. Each book presented unique layout challenges, often including margin notes in old lithographs, necessitating a highly adaptable processing solution.

Impact and Future Implications

The project's machine learning solution for image processing could benefit similar digitization efforts globally, particularly those dealing with complex scripts or varied source materials. This initiative highlights a community-driven approach to preserving cultural heritage, contrasting with large-scale commercial digitization efforts.