Sustav za automatsko očitavanje rukom pisanih ispravaka odgovora na ispitnim obrascima

master's thesis

• Title (english): System for Automated Recognition of Handwritten Corrections for Multiple Choice Answer Sheets
• Author: Domagoj Jakšić
• Mentor: Zoran Kalafatić (mentor)
• Committee member: Zoran Kalafatić (predsjednik povjerenstva)
• Committee member: Tomislav Hrkać (član povjerenstva)
• Committee member: Ante Đerek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U procesu ispunjavanja ispitnih obrazaca s pitanjima s više ponuđenih odgovora studenti ponekad pogriješe. Pogrešni unosi odgovora se mogu ispraviti upisivanjem slovne oznake željenog odgovora u za to predviđenu tablicu ispravaka. Rukom upisani ispravci odgovora otežavaju proces automatiziranog ocjenjivanja ispita jer ih operater treba ručno očitati te unijeti ispravljeni odgovor u postojeći sustav. U sklopu ovog diplomskog rada razvijen je sustav za automatsko očitavanje rukom pisanih ispravaka odgovora na ispitnim obrascima. U radu je pružen uvid u korištene skupove podataka za učenje i testiranje sustava. Potom su detaljno objašnjeni postupci normalizacije zakrenutosti ispitnog obrasca, detekcije tablice ispravaka u ispitnom obrascu te detekcije polja tablice ispravaka. U radu su predstavljeni modeli koji raspoznaju koja slovna oznaka je upisana u detektirano polje tablice ispravaka odgovora. Uz arhitekturu modela, navedeni su rezultati koje modeli postižu na testnom skupu ispitnih obrazaca.

Abstract (english)

In the process of filling out the exam sheets with questions with multiple choice answers, students sometimes make mistakes. Incorrect entries of the answers can be corrected by entering the letter label of the desired answer in the correction table. Handwritten answer corrections complicate the process of automated exam evaluation because the operator needs to manually review them and enter the corrected answer into the existing system. As a part of this graduate thesis, a system for automated recognition of handwritten corrections for multiple-choice answer sheets has been developed. The thesis provides an insight into the used data sets for training and testing the system. Furthermore, the procedures for normalizing the rotation of the exam sheet, detecting the answer correction table in the exam sheet, and detecting the answer correction table fields are explained in detail. Models that recognize which letter label is written in the detected field of the answer correction table are also presented in the thesis. In addition to the model architecture, the results that the models achieved on the test set of the exam sheets are listed.

• Keywords: računalni vid; raspoznavanje objekata; detekcija objekata; konvolucijske neuronske mreže; ispitni obrasci
• Keywords (english): computer vision; object recognition; object detection; convolutional neural networks; multiple-choice answer sheets
• Language: croatian
• URN:NBN: urn:nbn:hr:168:022218
• 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: 2023-01-13 12:29:41