Automatsko prepoznavanje rukom pisanih matematičkih izraza

master's thesis

• Title (english): Handwritten mathematical expression recognition
• Author: Nikola Vnučec
• Mentor: Marko Čupić (mentor)
• Committee member: Marko Čupić (predsjednik povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Committee member: Željka Mihajlović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-06-28, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom su radu prikazani načini rada i implementacija konvolucijskih i povratnih neuronskih mreža. Pomoću radnog okvira PyTorch razvijena su tri modela različitih složenosti za prepoznavanje rukom pisanih matematičkih izraza. Modeli su razvijeni u enkoder – dekoder arhitekturama. Dok je najjednostavniji model implementiran jednostavnim konvolucijskim i povratnim mrežama, najsloženiji model koristi složenu gusto povezanu neuronsku mrežu kao enkoder, a u ulozi dekodera su dvije jednosmjerne propusne povratne ćelije uz mehanizam pozornosti. Modeli su trenirani na javno dostupnom skupu podataka iz CROHME natjecanja koji sadrži nešto više od 10 000 slika. Na kraju rada su opisani dobiveni rezultati treniranja i testiranja modela.

Abstract (english)

In this work, the specific mechanisms and implementation details of convolutional and recurrent neural networks are shown. Three deep models of different complexity have been developed to recognize handwritten mathematical expression. Models are developed in encoder – decoder architectures. While the simplest model is implemented with basic structures of convolutional and recurrent neural networks, the most complex model uses DenseNet as encoder and multi-scale recurrent decoder with an attention mechanism. The models were trained on a public CROHME dataset containing just over 10 000 images. Following that, obtained numerical results of model training and testing are expressed and briefly discussed.

• Keywords: obrada slike; duboko učenje; konvolucijske neuronske mreže; povratne neuronske mreže; prepoznavanje matematičkih izraza
• Keywords (english): image processing; deep learning; convolutional neural networks; recurrent neural networks; mathematical expression recognition
• Language: croatian
• URN:NBN: urn:nbn:hr:168:598223
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-14 21:58:41