Optičko prepoznavanje rukopisa – usporedna analiza odabranih programskih rješenja

master's thesis

• Title (english): Handwritten text recognition – comparative analysis of selected software solutions
• Author: Josipa Sumpor
• Mentor: Hrvoje Stančić (mentor)
• Mentor: Željko Trbušić (pomoć pri izradi, neposredni voditelj)
• Committee member: Arian Rajh (predsjednik povjerenstva)
• Committee member: Željko Trbušić (član povjerenstva)
• Committee member: Hrvoje Stančić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Humanities and Social Sciences; (Department of information and Communication sciences); Zagreb
• Defense date and country: 2024-07-01, Croatia
• Scientific / art field, discipline and subdiscipline: SOCIAL SCIENCES; Information and Communication Sciences; Archives and Documentation Studies

Abstract

U ovom diplomskom radu istražuju se programska rješenja za optičko prepoznavanje rukopisa, s posebnim naglaskom na usporedbu rezultata platforme Transkribus i mobilnih aplikacija. Cilj je procijeniti točnost i pouzdanost različitih tehnologija za prepoznavanje rukopisa.
Analizirane su mobilne aplikacije OneNote, MyScript Nebo i Pen to Print, koje su se pokazale neučinkovitima. Kao napredna platforma, Transkribus nudi mogućnost treniranja vlastitih modela i podržava prepoznavanje velikog broja jezika i stilova rukopisa, dok mobilne aplikacije koriste jednostavnije algoritme i pružaju manje opcija za prilagodbu. Zbog ograničenja mobilnih aplikacija za prepoznavanje rukopisa, istraživanje se preusmjerilo na digitalizaciju gradiva I. klasične gimnazije u Zagrebu i treniranje vlastitog modela na Transkribusu. Digitalizacija je obuhvatila skeniranje i obradu 419 stranica, od kojih je 198 odabrano za daljnju izradu modela za prepoznavanje rukopisa. Provedena je, također, neovisna evaluacija rezultata koristeći ISRI evaluacijske alate.

Abstract (english)

This master thesis explores software solutions for handwritten text recognition, with a particular emphasis on comparing the results of the Transkribus platform and mobile applications. The aim is to assess the accuracy and reliability of various handwriting recognition technologies.
Mobile applications OneNote, MyScript Nebo, and Pen to Print were analyzed and found to be ineffective. Transkribus, as an advanced platform, offers the capability to train custom models and supports recognition across a wide range of languages and handwriting styles, whereas mobile applications employ simpler algorithms and provide fewer customization options. Due to the limitations of mobile handwriting recognition applications, the research shifted focus to digitizing materials from the First Classical Gymnasium in Zagreb and training a custom Transkribus model. The digitization process involved scanning and processing 419 pages, of which 198 were selected for further model development for handwriting recognition. Additionally, an independent evaluation of the results was conducted using ISRI evaluation tools.

• Keywords: optičko prepoznavanje rukopisa; Transkribus; mobilne aplikacije; digitalizacija; treniranje modela; OCR; HTR
• Keywords (english): handwritten text recognition; Transkribus; mobile applications; digitization; model training; OCR; HTR
• Language: croatian
• URN:NBN: urn:nbn:hr:131:122892
• Study programme: Title: Information Sciences; specializations in: Archivistics Study, Library Science, Information Sciences (research), Information Sciences (teaching), Museology and heritage management, Informatology; Course: Archivistics Study; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra informacijskih znanosti (magistar/magistra informacijskih znanosti)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Access restricted to authenticated users
• Created on: 2024-07-12 20:21:17