Duboki konvolucijski modeli za praćenje objekata

master's thesis

• Title (english): Deep convolutional models for object tracking
• Author: Ivan Fabijanić
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Zoran Kalafatić (član povjerenstva)
• Committee member: Leonardo Jelenković (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-19, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad se bavi problemom detekcije i praćenja sa dubokim konvoluciiskim modelima. Objašnjeni su koncepti dubokih konvolucijskih modela, metode učenja te metode detekcije i praćenja objekata. Također su opisani modeli detekcije i praćenja objekata koji su razvijeni u okviru ovog rada. Modeli su evaluirani tako da se mjerila kvaliteta praćenja modela za praćenje u odnosu na model za detekciju, mjerila se kvaliteta praćenja u odnosu na duljinu faze učenja i mjerila se brzina izvođenja ta dva modela. Testovi i faze učenja su provedeni na video snimci nogometne utakmice koja je snimana iz ptičje perspektive gdje su kao objekti za praćenje bili igrači oba tima i sudac.

Abstract (english)

This paper examines the problems of object tracking and detection with deep convolutional models. Concepts of deep convolutional networks, learning methods and methods for detection and tracking have been explained as well. Detection and tracking models developed for this paper have also been explained. Models are evaluated with measurement of tracking quality of tracking model against detection model, measurement of tracking quality with different length of learning phase and with measurement of execution performance of those two models. Learning phases and tests have been executed on video file containing the football match which have been filmed in bird's eye perspective where the tracking objects have been players of both teams and the referee.

• Keywords: praćenje objekata; detekcija objekata; konvolucijske neuronske mreže; računalni vid; duboko učenje; TensorFlow
• Keywords (english): object tracking; object detection; convolutional neural networks; computer vision; deep learning; TensorFlow
• Language: croatian
• URN:NBN: urn:nbn:hr:168:698674
• 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: 2019-03-13 20:10:30