Sustav za optičko raspoznavanje teksta starijih knjiga na hrvatskome jeziku

undergraduate thesis

• Title (english): Optical character recognition system for older books in Croatian
• Author: Dominik Agejev
• Mentor: Tomislav Hrkać (mentor)
• Committee member: Tomislav Hrkać (predsjednik povjerenstva)
• Committee member: Ivan Sabolić (član povjerenstva)
• Committee member: Ivan Martinović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Cilj rada nadići je uspješnost gotovih sustava za optičko raspoznavanje teksta na starijim knjigama hrvatskoga jezika koristeći se nenadziranim metodama učenja, uz predobradu i naknadnu obradu. Razmatraju se najznačajniji slobodno dostupni OCR alati prikladni zadatku, Tesseract, OCR sustav opće namjene koji održava Google, te Ocular, razvijen specifično za primjenu na antikvarnim dokumentima. Nakon treniranja i optimiziranja hiperparametara Oculara, uspoređen je s Tesseractom gdje se pokazuje da usprkos starijoj arhitekturi u bitnome nadjačava Tesseract, ali uz određena ograničenja. Konačno, izveden je sustav glasanja kojim se postiže veća uspješnost od one samostalnih modela.

Abstract (english)

The aim of the paper is to surpass the accuracy of out-of-the-box systems at Optical Character Recognition of historical documents in the Croatian language relying on unsupervised learning methods, preprocessing and postprocessing. The most appropriate freely available OCR tools are evaluated, namely Tesseract, a general-purpose OCR system maintained by Google, and Ocular, developed specifically for use on historical documents. After training and optimizing Ocular's hyperparameters it is compared to Tesseract where it is shown that despite its older architecture Ocular in the main still bests Tesseract, with certain caveats. Finally, a voting-based system is implemented which achieves greater success than each model alone.

• Keywords: OCR; optičko raspoznavanje teksta; računalni vid; Ocular; Tesseract
• Keywords (english): OCR; Optical Character Recogniton; Computer Vision; Ocular; Tesseract
• Language: croatian
• URN:NBN: urn:nbn:hr:168:755125
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 19:57:19