Praćenje lica pomoću dubokog učenja

master's thesis

• Title (english): Face Tracking Based on Deep Learning
• Author: Marko Krišković
• Mentor: Igor Sunday Pandžić (mentor)
• Committee member: Igor Sunday Pandžić (predsjednik povjerenstva)
• Committee member: Lea Skorin-Kapov (član povjerenstva)
• Committee member: Mirko Sužnjević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-10, Croatia

Abstract

Zadatak je bio napraviti algoritam praćenja lica koji radi na principu CNN-a i
usporediti ga s jednim od tradicionalnih algoritama praćenja lica. Model CNN-a koji sam
uzeo u ovu svrhu je YOLOv8, a software koji sam uzeo za uspoređivanje je Visage|SDK.
Označio sam bazu podataka za treniranje i istrenirao YOLOv8. Usporedio sam
performansu algoritama na slikama i snimkama. U ishodu uspoređivanja utvrdio sam da
oba algoritma imaju svoje nedostatke i prednosti te da se oba mogu specijalizirati... More za
različita područja. YOLO je bolji za praćenje detalja individualnih komponenti lica, a
Visage|SDK za praćenje oblika cijelokupnog lica. Less

Abstract (english)

The task was to create a face tracking algorithm that works on the CNN
principle and compare it with one of the traditional face tracking algorithms. The CNN
model I took for this purpose is YOLOv8, and the software I took for comparison is
Visage|SDK. I marked the training database and trained YOLOv8. I compared the
performance of algorithms on images and recordings. As a result of the comparison, I
found that both algorithms have their own advantages and disadvantages, and that both... More can
specialize in different areas. YOLO is better for tracking the details of individual facial
components, and Visage|SDK is better for tracking the shape of the entire face. Less

• Keywords: YOLO; CNN; duboko učenje; Visage|SDK; praćenje lica; strojno učenje
• Keywords (english): YOLO; CNN; deep learning; Visage|SDK; face tracking; machine learning
• Language: croatian
• URN:NBN: urn:nbn:hr:168:528009
• 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: 2023-10-11 09:26:12