diplomsko delo
Rok Pavlovič (Author), Vitomir Štruc (Mentor)

Abstract

Štetje množice je pomembna raziskovalna tema na področju računalniškega vida. Še vedno je težko natančno prešteti večje množice ljudi na festivalih, koncertih, protestih in zborih, kjer so ljudje natrpani skupaj. V zadnjih letih je štetje množic zelo napredovalo s pomočjo globokih nevronskih mrež. Metode globokega učenja so najsodobnejši pristop k štetju množic in oceni gostotne porazdelitve ljudi. V literaturi se pojavlja veliko postopkov z globokim učenjem, na njihovo uspešnost pa vpliva veliko faktorjev, kot so vremenske razmere, vrsta prizora, perspektiva in resolucija slike. V diplomski nalogi nas zanima kako se obneseta dve metodi štetja množic. Za našo analizo smo si izbrali CSRNet in MCNN. CSRNet (angl. Congested Scene Recognition Network) je metoda namenjena štetju ljudi v velikih gnečah, ki deluje na principu razširjene konvolucije. Druga metoda MCNN (angl. Multi-column Convolutional Neural Network) pa uporablja tri stolpične konvolucijske nevronske mreže, za boljšo razpoznavo različnih velikosti ljudi na sliki. Obe metodi smo ovrednotili na obeh delih podatkovne zbirke ShanghaiTech in na podatkovni zbirki UCF-CC-50. Eksperimente smo izvajali na na vseh treh zbirkah, pri čemer smo ločili podatke še na dodatne množice, da smo lahko analizirali vpliv zornega kota in vrste svetlobe. Naša analiza kaže, da se v povprečju boljše izkaže metoda CSRNet. Pri analizi zornega kota pridemo do zaključka, da imata modela boljše razultate pri nižjem zornem kotu. Pri faktorju vrste svetlobe pa lahko ugotovimo, da imata modela dobro razpoznavo ljudi tako v naravni svetlobi in pri umetno ustvarjeni svetlobi.

Keywords

štetje množic;strojno učenje;nevronske mreže;visokošolski strokovni študij;Aplikativna elektrotehnika;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FE - Faculty of Electrical Engineering
Publisher: [R. Pavlovič]
UDC: 004.932:004.85(043.2)
COBISS: 84942339 Link will open in a new window
Views: 263
Downloads: 23
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: Crowd-counting with machine learning methods
Secondary abstract: Crowd counting is an important research topic in the field of computer vision. It is still difficult to accuratley count larger crowds of people at festivals, concerts, protests and choirs, where people are crowded together. In the past years crowd counting has advanced greatly with the help of deep neural networks. Deep learning methods are the most modern approach to crowd counting and estimating human density. Many of them occur in literature and the estimations are influaced by many factors, such as weather conditions, scene type, perspective and image resolution. In the diploma thesis we are interested in how two crowd counting methods work. We selected CSRNet in MCNN for our analysis. CSRNet (Congested Scene Recognition Network) is a method designed to count people in large crowds, which works on the principle of dialated convolution. The second method MCNN (Multi-column Convolutional Neural Network) uses three column convolutional neural networks to better recognize the different sizes of people in the image. Both methods were evaluated on both parts of the ShanghaiTech dataset and on the UCF-CC-50 dataset. The experiments were performed on all three collections, separating the data into additional sets so that we could analyze the influence of the angle of view and the type of light. Our analysis shows that, on average, the CSRNet method performs better. In the analysis of the angle of view, we come to the conclusion that the models have better results at a lower angle of view. With regard to the type of light factor, we can conclude that the models have good recognition of people both in natural light and in artificially created light.
Secondary keywords: crowd counting;machine learning;neural network;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000315
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za elektrotehniko
Pages: 38 str.
ID: 13933685