magistrsko delo
Anže Bertoncelj (Author), Janez Perš (Mentor)

Abstract

Problem odčitavanja vrednosti analognih instrumentov z uporabo metod računalniškega vida je star problem, ki je bil že rešen z mnogimi različnimi pristopi. Zaradi zahteve podjetja po ponovni implementaciji starega algoritma, sta se nam pojavili dve vprašanji: Kako bi lahko problem rešili z uporabo najsodobnejših pristopov in kako dobro se ti novi pristopi primerjajo z obstoječimi starimi pristopi. V tej diplomski nalogi so predlagane tri metode, kjer vsaka poizkuša problem rešiti na svojevrsten način. Prva metoda je le ponovna implementacija in temelji na podlagi starih metod. Drugi dve metodi pa uporabljata umetno inteligenco in temeljita na nevronskih omrežjih VGG-16 in Mask R-CNN. V nalogi poleg opisov metod, te tudi implementiramo in med seboj primerjamo.

Keywords

nevronska omrežja;magisteriji;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UL FE - Faculty of Electrical Engineering
Publisher: [A. Bertoncelj]
UDC: 004.93:621.317.7(043.2)
COBISS: 107176195 Link will open in a new window
Views: 161
Downloads: 55
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Reading displays of measuring instruments using computer vision and machine learning methods
Secondary abstract: The problem of reading values from analogue instruments using comptures vision methods is an old problem that has already been solved many times using various methods. A request to re-implement the algorithm led to us ask ourself two questions: how could the problem be solved using state-of-the-art approaches, and how well do these new approaches compare with the existing old ones. In this thesis, three methods are presented, each trying to solve the problem of in a different way. The first method is just a re-implementation and upgrade of existing methods. However, the other two methods use artificial intelligence and are based on two different neural networks VGG-16 and Mask R-CNN. In addition to describing the methods, we also implement them and compare their results.
Secondary keywords: analogue measuring instruments;computer vision;machine learning;neural networks;computer science;master's degree;Merilni instrumenti;Računalniški vid;Strojno učenje;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Master's thesis/paper
Study programme: 1001017
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 90 str.
ID: 15129360