Search the digital library catalog

Query:	search in
	search in
	search in
	search in
Work type:	* old and bologna study programme
Language:
Search in:
Options:	Show only hits with full text
	Reset

1 - 2 / 2

An analysis of exploration and exploitation using attraction basins on 2D and 3D continuous functions : master's thesis
Mihael Baketarić, 2020, master's thesis

Abstract: In this thesis we were discussing an analysis of numerical optimization algorithms from the most important aspect, that is exploration and exploitation. We focused on 2-dimensional and 3-dimensional unconstrained continuous functions, which were used to test the recently proposed metric based on attraction basins. The metric does not need any user-defined parameters. Attraction basins were expounded more profoundly and extensively. Our algorithm to calculate them consists of three steps such as making potential boundaries, filling, and then removing false boundaries from attraction basins. Results show that our algorithm is barely satisfying, depends on a particular problem function used. For example, attraction basins from Rastrigin, Schwefel, Ackley and similar functions (including all unimodal ones) were calculated accurately, while more special functions like Michalewicz, Shubert and Branin were proved to be not so easy. Further, we arbitrarly selected two algorithms, Particle Swarm Optimization and Self-adapting Differential Evolution, not for comparative study, rather to test the metric based on attraction basins. Results implied the relevance of recently proposed metric, and opened us a fruitful field for further investigation.
Keywords: exploration, exploitation, attraction basins, optimization, metaheuristic
Published in DKUM: 04.11.2020; Views: 887; Downloads: 93
Full text (1,76 MB)

Prepoznavanje aktivnosti osebe iz zaporedja slik s pomočjo konvolucijskih nevronskih mrež
Mihael Baketarić, 2018, undergraduate thesis

Abstract: V diplomskem delu smo se ukvarjali s prepoznavanjem aktivnosti osebe iz zaporedja slik. Omejili smo se na aktivnosti: stoji, sedi, leži, hitro hodi, počasi hodi in pada. Pregledali smo obstoječe postopke prepoznavanja, pripravili množico podatkov, preučili konvolucijske nevronske mreže in jih uporabili pri reševanju našega problema. Naš algoritem je sestavljen iz dveh korakov: iz izločevanja oseb iz slik in prepoznavanja aktivnosti. Oba koraka smo implementirali z uporabo konvolucijskih nevronskih mrež in analizirali rezultate. Za učenje in testiranje smo uporabili lastno podatkovno zbirko, ki je vsebovala video posnetke 6-ih različnih oseb, ki so izvajali vseh šest aktivnosti. Na veliko slikah oseba ni bila pravilno izločena oz. detektirana, zato se je naša množica podatkov občutno zmanjšala po odstranitvi takšnih slik. Naš postopek smo preverili s 6-kratno navzkrižno validacijo. Povprečna uspešnost prepoznavanja aktivnosti je bila 36 %, kar seveda ni dovolj visoko za realne aplikacije. Ugotavljamo, da se pri rezultatih prepoznavanja aktivnosti močno pozna dejstvo, da v našem postopku nismo upoštevali časovne komponente oz. rezultatov prepoznav na predhodnih slikah.
Keywords: računalniški vid, konvolucijska nevronska mreža, globoko učenje, detekcija oseb, prepoznavanje aktivnosti osebe
Published in DKUM: 19.10.2018; Views: 2088; Downloads: 262
Full text (1,55 MB)

Search done in 9.76 sec.