Your browser does not allow JavaScript!
JavaScript is necessary for the proper functioning of this website. Please enable JavaScript or use a modern browser.
Repository of the University of Ljubljana
Open Science Slovenia
Open Science
DiKUL
slv
|
eng
Search
Browse
New in RUL
About RUL
In numbers
Help
Sign in
Details
Izbor estetsko optimalne fotografije iz sekvence podobnih ali skorajda enakih fotografij
ID
HOŽIČ, TOMAŽ
(
Author
),
ID
Solina, Franc
(
Mentor
)
More about this mentor...
PDF - Presentation file,
Download
(15,44 MB)
MD5: B3AE764630488E2364B08C35C85B92BB
PID:
20.500.12556/rul/d890b380-4bac-4e22-a120-a07991c710a2
Image galllery
Abstract
Digitalna fotografija je povzročila eksponentno povečanje obsega posnetih fotografij. Pregledovanje in na podlagi estetskih kriterijev utemeljen izbor pravih, najboljših oz. najlepših posnetkov zahteva precej časa. Z metodami strojnega učenja so že dosegli odlične in ponovljive rezultate prepoznavanja obrazov in objektov ter kategorije fotografij (portret, pokrajina, makro posnetek, šport idr.). Zaradi vse bolj zmogljive strojne opreme so za iskanje učinkovitih in uspešnih rešitev ponovno postale zelo zanimive nevronske mreže. Magistrsko delo analizira obstoječe rešitve oz. orodja, ki omogočajo (pol)avtomatski izbor najbolj estetskih fotografij iz sekvence fotografij. V nadaljevanju primerja klasični pristop za estetsko ugotavljanje in ocenjevanje s pomočjo metod in orodij strojnega učenja ter trend iskanja estetskih ocen oz. klasifikacij s pomočjo metod in orodij globokih konvolucijskih nevronskih mrež (angl. Deep Convolutional Neural Networks). Na podlagi lastnih profesionalnih fotografskih izkušenj, lastnega slikovnega materiala in preizkusa komercialnega prototipa je podan predlog rešitve problema.
Language:
Slovenian
Keywords:
fotografija
,
estetika
,
estetika fotografije
,
strojno učenje
,
konvolucijske nevronske mreže
,
metapodatki
Work type:
Master's thesis
Organization:
FRI - Faculty of Computer and Information Science
Year:
2016
PID:
20.500.12556/RUL-85899
Publication date in RUL:
28.09.2016
Views:
2239
Downloads:
464
Metadata:
Cite this work
Plain text
BibTeX
EndNote XML
EndNote/Refer
RIS
ABNT
ACM Ref
AMA
APA
Chicago 17th Author-Date
Harvard
IEEE
ISO 690
MLA
Vancouver
:
HOŽIČ, TOMAŽ, 2016,
Izbor estetsko optimalne fotografije iz sekvence podobnih ali skorajda enakih fotografij
[online]. Master’s thesis. [Accessed 2 April 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=85899
Copy citation
Share:
Secondary language
Language:
English
Title:
Choosing an aesthetically optimal photographic image from a sequence of similar or nearly identical images
Abstract:
Digital Photography has caused an exponential increase of captured images. The process of reviewing and selecting the most beautiful images, in regard to some well-known aesthetic criteria, is a time consuming task. Excellent and repeatable face and object-recognition and image classification tasks of various kinds of images (portrait, landscape, macro, sports) are the result of applied machine learning algorithms. Because of the increase in hardware computing-power deep neural networks are becoming more and more popular. This master thesis analyses solutions and tools which can (semi)automatically find the most aesthetically pleasing images from a sequence of images. In this thesis the methods and tools of classical machine learning and those of deep convolutional neural networks are compared. On the basis of my professional photographic experience and images in various fields of photography (documentary, wedding, sport) a comercial prototype application is tested and some solutions to the problem are suggested.
Keywords:
photography
,
aesthetics
,
aesthetics of photography
,
machine learning
,
convolutional neural networks
,
metadata
Similar documents
Similar works from RUL:
Searching for similar works...
Similar works from other Slovenian collections:
Back