Prepoznavanje osoba na temelju fotografije očiju

master's thesis

• Title (english): Identifying People Based on Eye Photograph
• Author: Katarina Šikić
• Mentor: Marko Subašić (mentor)
• Committee member: Marko Subašić (predsjednik povjerenstva)
• Committee member: Sven Lončarić (član povjerenstva)
• Committee member: Zvonko Kostanjčar (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom se radu razmatra problem prepoznavanja osoba na temelju fotografije očiju. U tu je svrhu obrađena tema pronalaska lica na fotografijama, detekcije očiju, izdvajanja područja na kojemu se nalaze oči i augmentacije fotografija. Opisane su teme umjetnih, dubokih i konvolucijskih neuronskih mreža i njihova učenja. Za detekciju očiju korištena je konvolucijska neuronska mreža YOLOv3 dotrenirana fotografijama pripremljenim korištenjem algoritma dlib. Za prepoznavanje osobe korištena je konvolucijska neuronska mreža implementirana u radnoj okolini TensorFlow. Za njezino su učenje korištene fotografije očiju modificirane korištenjem biblioteke OpenFace kako bi odgovarale potrebama rješavanja ovoga problema. Na kraju su prikazani rezultati testiranja detektora i klasifikatora na dvije baze podataka.

Abstract (english)

This paper discusses the problem of recognizing persons based on eye photography. For this purpose, the topic of finding faces in photographs, eye detection, extracting the area where the eyes are located and augmenting photographs is covered. The topic of artificial, deep and convolutional neural networks and their learning is described. A convolutional neural network YOLOv3 trained with photographs prepared using the dlib algorithm was used for eye detection. A convolutional neural network implemented in the TensorFlow work environment was used to identify the person. Eye photographs modified using the OpenFace library to meet the needs of solving this problem were used for training. Finally, the results of testing detectors and classifiers on two databases are presented.

• Keywords: klasifikacija; konvolucijske neuronske mreže; duboko učenje; detekcija objekata
• Keywords (english): classification; convolutional neural networks; deep learning; object detection
• Language: croatian
• URN:NBN: urn:nbn:hr:168:214521
• 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 13:55:30