Search

This master's thesis deals with the problem of selecting features for automatic detection of objects of a given visual category in images. The goal of object detection is to determine the locations and sizes of all objects of the given category in each input image, where the number of objects displayed in individual images is not known in advance. Various circumstances may contribute to the difficulty of object detection, such as cluttered backgrounds, diversity in the appearance of objects in the category, diversity in the scale of the displayed objects, and partial occlusions. The thesis primarily focuses on systems that represent each input image by a set of features and use the resulting image representations to form detection hypotheses. The feature set is assumed to be prepared by a two-step procedure in the training stage. In the first step, the initial feature set is extracted from a set of training images. In the second step, the features to be used in the test stage are selected from the initial set. The thesis is devoted to the second step, i.e., feature selection. An effective selection method reduces the computational complexity in the test stage and eliminates features that are useless or even harmful for detecting objects of a given category. In the master’s thesis, we present two feature selection methods. At the core of the first one (the so-called filter method) is a feature evaluation function that is based on a transformation of the problem of detecting objects in training images into a problem of classifying training images themselves. Since much existing theory and practice of feature selection pertains to the classification rather than detection domain, such a problem transformation greatly expands the range of applicable selection approaches. Because of its design, the feature evaluation function can be straightforwardly incorporated into the AdaBoost selection framework, which selects features by considering their interdependence in an implicit manner. The second selection method that is presented in this thesis (the so-called wrapper method) systematically runs the detection procedure on a fixed training image set for various candidate feature sets. The goal of such a selection approach is to find a feature set that enables the detection procedure to attain the highest detection accuracy on the training image set. The problem of finding the optimal candidate feature set can be easily transformed into a problem of finding the optimal vertex in the corresponding state space graph. In the implementation of the system, each of the two selection methods was integrated with a well-known detection approach and with two promising state-of-the-art feature extraction methods. We experimentally tested all four combinations of feature extraction and feature selection methods. The experiments were evaluated using five test image sets, each of which defined a visual category and an independent detection setup. In some cases, as it turned out, a very small number of selected features suffices to bring about a fairly high detection accuracy. Particularly remarkable results were achieved for the combination of the filter selection method, the feature extraction method due to Fidler et al. (2006), and the UIUC car dataset. In this setup, the selection of merely four features out of 952 initial candidates led to a detection accuracy that is comparable to the state-of-the-art results for the UIUC dataset.

A graph grammar is a generative description of a graph language (a possibly infinite set of graphs). In this paper, we present a novel algorithm for inducing a graph grammar from a given set of 'positive' and 'negative' graphs. The algorithm is guaranteed to produce a grammar that can generate all of the positive and none of the negative input graphs. Driven by a heuristic specific-to-general search process, the algorithm tries to find a small grammar that generalizes beyond the positive input set. During the search, the algorithm employs a graph grammar parser to eliminate the candidate grammars that can generate at least one negative input graph. We validate our method by inducing grammars for chemical structural formulas and flowcharts and thereby show its potential applicability to chemical engineering and visual programming.

Generics are a powerful tool in object-oriented programming languages, especially in the Java programming language. They play an important role in software development, since they enable the programmer to write typeindependent code in a type-safe way. In a nutshell, a generic Java class (or method) is a class (or method) parameterized with an arbitrary number of types. When using a generic class or method, we specify concrete values for the type parameters, effectively tailoring the code of the class or method for concrete types. Another benefit of Java generics is that type safety is enforced during compilation rather than execution. The C++ programming language has a concept parallel to Java generics these are templates. C++ templates and Java generics are similar to one another in terms of code syntax, the provided support for parameterizing types, and the most frequent use cases. However, they are actually implemented in a significantly different way, and there are important differences in their use and background semantics. In this thesis, we will show examples for both Java generics and C++ templates. We will discuss their similarities and differences, as well as advantages and disadvantages. Generiki so močno orodje v objektno usmerjenih programskih jezikih, še zlasti v jeziku java. Pri razvoju programske opreme igrajo pomembno vlogo, saj omogočajo programerju, da piše kodo, ki je neodvisna od konkretnih tipov, a kljub temu jamči pravilno rabo tipov. Generični javanski razred (oz. metoda) je v osnovi razred (oz. metoda), ki je parametriziran(a) s poljubnim številom tipov. Pri uporabi generičnega razreda oz. metode podamo konkretne vrednosti za tipne parametre in tako prilagodimo kodo razreda oz. metode za konkretne tipe. Druga prednost javanskih generikov je, da se skladnost tipov preverja že v času prevajanja, ne šele v času izvajanja. Koncept, podoben generikom v javi, obstaja tudi v jeziku C++: to so t.i. predloge. Predloge v C++ in generiki v javi so si med seboj podobni po sintaksi, po podpori parametrizaciji tipov in po najpogostejših primerih uporabe, močno pa se razlikujejo po implementaciji. Bistvene razlike lahko najdemo tudi pri uporabi in semantiki generikov in predlog. V tej diplomski nalogi se bomo ukvarjali tako z generiki v javi kot s predlogami v C++. Govorili bomo o njihovih medsebojnih podobnostih in razlikah, pa tudi o prednostih in slabostih.

V današnjem svetu ima splet ogromno vlogo. Veliko vidikov življenja se prepleta z uporabo spletnih aplikacij. Eden izmed njih je naročanje na termine storitev. Izkaže se, da veliko ponudnikov storitev uporablja zastarele dolgotrajne tehnike naročanja ali pa sploh ne ponujajo možnosti neverbalnega naročanja. Čeprav podobne rešitve že obstajajo, smo se odločili za implementacijo lastne, saj smo želeli razviti enostaven in unikaten uporabniški vmesnik. Poskrbeli smo, da se spletna aplikacija na vseh napravah pravilno obnaša in zagotavlja konsistentno uporabniško izkušnjo. Z uporabo sodobne zbirke tehnologij MEAN smo razvili spletno aplikacijo, katere uporaba je preprosta in intuitivna ne glede na izkušnje uporabnika s spletnimi aplikacijami. In today's world, the web plays a huge role. Many aspects of life are intertwined with the use of web applications. One of these is booking service appointments. It turns out that many service providers use outdated, time-consuming booking techniques or do not offer a non-verbal booking options at all. Although similar solutions already exist, we decided to implement our own because we wanted to develop a simple and unique user interface. We've made sure the web app behaves correctly on all devices and provides a consistent user experience. Using MEAN's modern suite of technologies, we have developed a web application that is simple and intuitive to use, regardless of the user's experience with web applications.

Slack je platforma za komunikacijo, katere glavni namen je izboljšati sodelovanje znotraj organizacij. Platforma je zlasti priljubljena med podjetji, ki se ukvarjajo z razvojem programske opreme, dnevno pa jo uporablja kar 12 milijonov posameznikov. Razvijalci lahko v platformi Slack izdelujejo lastne aplikacije, s katerimi rešujejo probleme uporabnikov ter optimizirajo in avtomatizirajo vsakodnevne delovne procese. V diplomski nalogi bomo opisali postopek izdelave aplikacije za platformo Slack ter predstavili primer aplikacije za pošiljanje anket in prikazovanje njihovih rezultatov. Ogledali si bomo, kaj nam platforma Slack nudi z vidika uporabnika in razvijalca, ter predstavili sodobne tehnologije, ki nam bodo pomagale pri razvoju. Slack is a communication platform designed to enhance cooperation in various organizations. With 12 million daily users, Slack is one of the most popular communication platforms today, especially among software developers. To solve common problems, as well as simplify and automate the everyday tasks of their users, Slack allows developers to create their own applications. In the thesis, we describe how to develop an application for Slack and present a sample application for sending surveys and displaying their results. We discuss Slack's capabilities from the user's and developer's perspective and describe modern technologies used in the development process.

Sestava optimalne ekipe ob upoštevanju podanih omejitev je zahteven problem celo za vešče in izurjene oči, zato njegova rešitev za športne institucije predstavlja veliko dodano vrednost. Pokazali bomo, da lahko različico problema z eno samo omejitvijo preslikamo na bolj znani problem nahrbtnika (angl. Knapsack problem), za katerega že obstajajo metode reševanja. Predstavili bomo metodo za ocenjevanje kakovosti športnika in celotne ekipe, uvedli pristop linearnega programiranja in predstavili lasten program v programskem okolju Python, ki iz podatkovne množice več deset tisoč igralcev sestavi optimalno ekipo. Zaznali bomo nekatere zakonitosti, ki jih bežni spremljevalec športa morda ne bi spoznal, med njimi npr. prenizko vrednotenje izkušenih igralcev ali previsoko vrednotenje igralcev v EPL. Pridobljene rezultate bomo ocenili ter primerjali z obstoječima, v svetu športa konkurenčnima ekipama. Ugotovili bomo, da so rešitve programa smiselne in da je kakovost sestavljene ekipe odvisna predvsem od števila in strogosti podanih omejitev (manj omejitev nam omogoča sestavo boljše ekipe in obratno). Since the problem of forming an optimal team subject to given constraints is demanding even for skilled and trained eyes, a solution to it represents a great added value to sports institutions. We show that the single-constraint version of the problem is equivalent to the better known knapsack problem, for which there are already methods to solve it. We present a method for assessing the quality of a single player and the entire team, introduce a linear programming approach, and present our own program written in the Python software environment that forms an optimal team from a data set of tens of thousands of players. We evaluate the obtained results and compare them with the existing teams competing in the world of sports. We find out that the results produced by the program make sense and that the quality of the composed team depends mainly on the number and severity of the given constraints (fewer constraints enable us to form a better team and vice versa). Furthermore, we discover a few rules that a casual sports fan might now know, among them, e.g., underestimation of experienced players or overestimation of players in the EPL.

In this thesis, we present a novel sports team management application called VTeamTrack. We focus on its structure, architecture, benefits, and goals. In order to help a team organize their time or improve their schedule, and to enable quick access to team-related information, the VTeamTrack application provides the following functionalities: game and training attendance overview, posts feed, creating and taking surveys, overview of membership payments, access to public teams’ schedules, and more. The application supports five user roles: player, coach, administrator, supervisor, and guest. The architecture of the application is based on the Android application architecture standard. Such an application is composed of two mandatory layers (user interface layer and data layer) and one optional layer (the domain layer). V tem diplomskem delu je opisan razvoj mobilne aplikacije za upravljanje športne ekipe. Aplikacija, imenuje se VTeamTrack, ni namenjena samo igralcem in trenerju, ampak tudi splošnim uporabnikom (gostom). Aplikacija omogoča komunikacijo med igralci in pomaga trenerju pri enostavnem upravljanju, koordinaciji in organiziranju ekipe in njenih članov. Športne ekipe se od drugih konvencionalnih skupin razlikujejo v več pogledih. Od svojih članov med drugim terjajo resnost, odgovornost in trdo delo. Člani morajo redno trenirati, upoštevati navodila trenerja in tako na tekmah kot na treningih dati vse od sebe. Aplikacija VTeamTrack je plod opazovanja izzivov, s katerimi se vsak dan srečujejo člani športne ekipe. Aplikacijo smo zasnovali tako, da lajša tovrstne izzive --- predvsem v organizacijskem smislu. Uporabnik aplikacije (igralec ali trener) lahko pregleda urnik treningov in tekem ter seznam vseh igralcev ekipe. Igralci lahko označijo, ali se nameravajo udeležiti treninga. Njihova namera je vidna vsakemu igralcu. Na ta način je mogoče preprečiti marsikatero komunikacijsko in organizacijsko težavo. Da bi bila ekipa na tekočem z morebitnimi spremembami, posodobitvami ali novicami, smo implementirali funkcionalnost objavljanja in komentiranja. Osebni podatki uporabnikov so dostopni prek njihovih profilov, kar omogoča enostavno in preprosto pridobivanje kontaktnih podatkov trenerja ali novega igralca. Igralec lahko tudi preverja stanje plačila članarine in je opozorjen, če je za tekoči mesec še ni plačal. Ko je članarina plačana, trener označi plačilo v aplikaciji in mu zanj ni več treba skrbeti. Zapisi o plačilih se vodijo skozi celotno leto. Poleg naštetih funkcionalnosti aplikacija ponuja še možnost ustvarjanja in izvajanja anket. Na ankete lahko odgovarja katerikoli igralec, ustvarja pa jih lahko samo igralec z administratorskimi pravicami. V aplikaciji poleg trenerja in igralca nastopata še dve uporabniški vlogi: gost in nadzornik. Gost lahko ob vstopu na začetni zaslon aplikacije pregleda razpored iger in treningov tistih ekip, ki so se odločile za javno dostopnost teh podatkov. Seveda lahko ekipe status vidnosti omenjenih podatkov kadarkoli spet nastavijo na “zasebno”. V vlogi nadzornika lahko nastopa starš ali skrbnik igralca. Uporabo te možnosti priporočamo mlajšim ekipam, ker ustvarja občutek varnosti za igralca in pomaga staršem, da so na tekočem. Nadzornik si lahko ogleda status plačila članarine in udeležbo igralca na igrah in treningih. Tako nadzornikovo kot igralčevo registracijo mora potrditi trener, ta pa je poleg tega edini, ki lahko iz aplikacije odstrani kateregakoli uporabnika. Trener lahko izbranemu igralcu dodeli administratorske pravice. Te pravice vključujejo ustvarjanje anket, preverjanje in beleženje prisotnosti na treningih, preverjanje prisotnosti na tekmah in drugo. Trener lahko upravlja nastavitve aplikacije. Nastavitve vključujejo nadzor statusa podatkov o ekipi (javni ali zasebni) in nastavitev časa za shranjevanje zapisov treningov in tekem. Vse funkcionalnosti in njihov potek so podrobno opisane z uporabniškimi tokovi, nekatere pa smo opisali tudi z diagrami ali shemami. Drugi del diplomske naloge razlaga arhitekturo aplikacije, saj je ta ključnega pomena za razvojni proces. Aplikacija je strukturirana v skladu s tipičnimi arhitekturnimi vzorci aplikacij za sistem Android. Razdeljena je na tri glavne plasti, pri čemer je uporabniški vmesnik karseda ločen od podatkovne plasti in izvajalne logike. Arhitektura aplikacije temelji na dveh glavnih načelih: Ločevanje zadolžitev. Vsaka komponenta aplikacije ima svojo točno določeno zadolžitev. Ločevanje uporabniškega vmesnika od podatkovnih modelov. To med drugim zagotavlja, da se podatki aplikacije ne izgubijo, če operacijski sistem uniči proces aplikacije. Aplikacijo smo razvili v integriranem razvojnem okolju Android Studio. Uporabljali smo objektno orientiran programski jezik java in ogrodje Android SDK (Software Development Kit). Tretji del diplomske naloge poskuša opredeliti značilnosti aplikacije, ki so pomembne za optimalno izkušnjo njenega uporabnika. Uspeh aplikacije je neposredno povezan s stopnjo vključenosti uporabnika. Analiza pričakovanj uporabnikov in njihova vključenost pred (in med) razvojnim procesom sta ključnega pomena za izdelavo uspešnega izdelka, saj je pri vsaki načrtovani interakciji z izdelkom ali storitvijo treba identificirati potrebe uporabnikov. Aplikacija VTeamTrack je namenjena predvsem manjšim, mlajšim in lokaliziranim ekipam, zato smo jo oblikovali tako, da se prilagaja njihovim potrebam. Z realizacijo omenjenih funkcionalnosti smo cilj našega projekta dosegli: opredelili in razvili smo aplikacijo, ki bo športni ekipi služila kot organizacijsko orodje, načrtovalec, opomnik in način komunikacije, ki poskrbi, da so vsakemu članu ekipe vse potrebne informacije takoj dostopne.

V diplomski nalogi predstavljamo, kako lahko z enim naborom kode testiramo grafične vmesnike mobilnih aplikacij tako za platformo iOs kot za platformo Android. Opisani pristop k testiranju primerjamo s testiranjem z uporabo domorodnih ogrodij za omenjeni platformi. Namen dela je pokazati prednosti in slabosti takšnega testiranja ter s tem pomagati pri izbiri orodja za testiranje grafičnih vmesnikov mobilnih aplikacij. Avtomatizacija testiranja programske opreme je danes izrednega pomena, saj znižuje stroške in pospešuje proces testiranja. V okviru tega diplomskega dela obravnavamo delovanje, arhitekturo in implementacijo testov z orodjema Appium in Pytest. Implementirane teste nato primerjamo z domorodnimi testi za platformi iOs in Android. Dobljene rezultate analiziramo in komentiramo. In this thesis, we present how to test the user interface of mobile applications for both the iOs and Android platform using a single code base. We compare the described approach to testing with approaches that use native frameworks for iOs and Android. The purpose of this work is to present the advantages and disadvantages of such testing and help the reader decide about the proper tool for the automation of user interface testing for mobile applications. Nowadays, automation of software testing is of great importance, since it reduces the cost and speeds up the process of testing. In this thesis, we present the behavior, architecture, and implementation of automated tests with the tools Appium and Pytest. Subsequently, we compare automated single-code-base tests with tests for the native frameworks for each platform and comment on the results.

Ker se svet pospešeno digitalizira, se vedno več datotek shranjuje na digitalnih medijih. Ti so podvrženi okvari in izgubi podatkov, zato obstaja čedalje več ponudnikov storitev oblačnega shranjevanja datotek. Ker njihove programske rešitve praviloma niso dostopne javnosti, jim marsikdo ne zaupa. Da bi omilili opisani problem, smo razvili odprtokodni sistem mnimidamon, ki uporabnikom omogoča, da s svojimi računalniki tvorijo skupine za varnostno kopiranje datotek in tako sami ustvarijo "oblake", ki bodo hranili varnostne kopije njihovih datotek. Uporabniki tako ohranijo nadzor nad svojimi datotekami in jim ga ni več treba predati v roke zunanjemu ponudniku. Pri razvoju sistema se nismo ukvarjali zgolj z varnostnim kopiranjem datotek, ampak smo velik poudarek namenili tudi preprosti uporabniški izkušnji vzpostavitve in uporabe sistema. With the world digitizing itself in an ever-increasing pace, more and more files tend to be stored on digital media. Since these are liable to failures and data losses, there has been a surge in the number of services providing cloud data storage. Unfortunately, the source code of their solutions is usually not accessible to the general public as a result, many people tend to distrust such services. To alleviate this problem, we developed an open-source system called mnimidamon that enables the users to organize their computers into file backup groups and thus themselves create "clouds" to store backups of their files. In this way, the users maintain control over their files and do not have to entrust it to an external provider. During the development of our system, we did not restrict ourselves just to the problem of creating backups, but we also tried our best to ensure a simple setup and a pleasant user experience.

Na fakulteti se je skozi leta uveljavila uporaba testnih primerov za samodejno preverjanje delovanja študentskih programov. Za njihovo pripravo so odgovorni bodisi izvajalci bodisi študentje. Študentje si testne primere med seboj pogosto izmenjujejo. Zaradi želje po večjem nadzoru nad kakovostjo in količino ter bolj pravični delitvi bremena izdelave testnih primerov med študenti smo v okviru diplomske naloge implementirali storitev, ki omogoča izmenjevanje testnih primerov določene naloge, pri čemer je vsak testni primer lahko avtomatsko validiran glede na pravilnost in podobnost. Uporabnik sistema za vsak oddan testni primer pridobi določeno število žetonov, ki jih lahko porabi za nakup testnih primerov drugih uporabnikov. Izvajalec predmeta pa pridobi vpogled v testne primere študentov, kar mu omogoča spremljanje razvoja znanja pri predmetu. Over the years, test cases have become an established tool to automatically verify students' programs at our faculty. Test cases are prepared either by the course provider or by individual students. The students often exchange their test cases. To enable a greater control of quality and quantity and to achieve a fairer division of the burden of preparing test cases among students, we implemented a service that makes it possible to exchange the test cases associated with a particular programming problem, where each test case can be automatically validated for correctness and similarity. For each test case submitted, the user of the system receives a certain number of tokens that can be used to purchase test cases from other users. The course provider gains insight into the students' test cases, which allows him to monitor the development of knowledge in the course.

V okviru diplomske naloge smo s pomočjo orodja Unity izdelali igro za predstavitev delovanja različnih algoritmov za tvorbo labirintov in iskanje poti v njih. Igra ponuja različne igralne načine in z vsakim nam poiskuša prikazati delovanje posameznih algoritmov. Eden izmed igralnih načinov nam omogoča tvorbo labirintov z devetimi različnimi algoritmi. Drugi nam omogoča reševanje labirintov s šestimi različnimi algoritmi in njihovimi kombinacijami. Tretji igralni način pa je namenjen tistim, ki si želijo razmigati možgane, saj vnaprej pripravi naključne labirinte in izriše pot, ki ji morajo uporabniki slediti s pomočjo sestavljanja iskalnih algoritmov, da uspešno rešijo labirint. Igra je zasnovana tako, da nam omogoča čim večji nadzor nad izvajanjem algoritmov (upočasnjevanje, ustavljanje, premik za en korak, izpis podatkov med izvajanjem algoritmov). In this thesis, we developed a Unity-based game to visualize different algorithms for generating labyrinths and finding paths through them. The game offers different modes to present the algorithms in different ways. One of the game modes allows us to generate labyrinths using nine different algorithms. Another enables us to solve labyrinths with six different algorithms and their combinations. The last game mode is intended for users who are looking for some brain stimulation, as it prepares random labyrinths in advance and plots the path that users have to follow by combining search algorithms in order to successfully solve the labyrinth. The game is designed to give us as much control over the execution of algorithms as possible (slowing down, stopping, step-by-step execution, showing relevant data during execution).

V diplomski nalogi predstavljamo interaktivno aplikacijo, ki vizualizira delovanje nekaterih algoritmov za iskanje najkrajših poti na grafih. Iz precej širokega nabora algoritmov smo izbrali Dijkstrov, Bellman-Fordov, Floyd-Warshallov, D'Esopo-Papeov in Fredman-Tarjanov algoritem. V besedilu teoretično predstavljamo njihovo delovanje in uporabo na različnih področjih, navajamo pa tudi njihovo psevdokodo. Predstavljamo tudi delovanje aplikacije in primere delovanja za posamezne algoritme. Aplikacijo smo izdelali kot didaktički pripomoček, namenjen tako učiteljem kot študentom. In this diploma thesis, we present an interactive application for visualizing algorithms for finding shortest paths in graphs. From a fairly large set of algorithms, we selected the Dijkstra, Bellman-Ford, Floyd-Warshall, D'Esopo-Pape, and Fredman-Tarjan algorithm. We describe them informally and in pseudocode. Tha application was designed as a didactic tool, both for teachers and for students.

Pri razvoju programske opreme razvijalec poleg zadanih funkcionalnosti ustvari tudi napake, t. i. hrošče. Njihovo iskanje in odpravljanje, imenovano tudi razhroščevanje, sodi med dražje in dolgotrajnejše procese, zato raziskovalci razvijajo samodejne metode lokalizacije napak. Cilj tega magistrskega dela je primerjati množice metričnih spektralnih metod lokalizacije napak v programu in med njimi določiti najboljšo. V ta namen smo razvili repozitorij načrtno pokvarjenih testnih programov in ga vodili v aplikacijo FaultLocalization, ki je na izhodu vrnila urejen rangiran seznam metod. Z izvedeno analizo rezultatov smo pokazali, da osrednja referenčna metoda Tarantula zagotavlja zelo dobre rezultate, kot najboljša pa se je pokazala metoda Overlap. Software development is a process in which a developer produces specified functionality as well as software errors, called bugs. The process of finding and fixing those errors, called debugging, is one of the most expensive and time-consuming parts of software development which is the main reason why many automatic fault localization methods have been developed. The main goal of this master thesis is to compare multiple software fault localization methods and to determine which provides the best results. We have developed a repository of faulty test programs and led it into the application FaultLocalization, which returns a ranking list of chosen methods as its output. The results of the analysis performed as part of the thesis prove that the main reference method Tarantula provides very good results, while the best results are achieved by the Overlap method.

tnost programskih konceptov in pomanjkanje interesa študentov sta glavna vzroka za visoko stopnjo neuspeha na uvodnih tečajih programiranja in visoke stopnje osipa študentov na fakultetah. Da bi pomagali študentom in jih motivirali, raziskovalci nenehno predlagajo različna orodja. Ena od številnih takšnih metod je vizualizacija algoritmov, ki izboljšuje kakovost pouka in lajša učenje algoritmov. Človeški možgani namreč lažje obdelajo vizualne slike kot obsežne kose izvorne kode, zato človek lažje razume postopek izvajanja algoritmov. Cilj je najti različne pristope in znižati raven abstrakcije algoritmov. Ciljna skupina so učenci, ki jih zanima področje računalništva in informatike in nimajo predznanja s področja algoritmov urejanja ali pa se teh algoritmov ravno učijo. Spletna aplikacija, izdelana v diplomski nalogi, vključuje učenje po korakih (to je eden od najučinkovitejših načinov razumevanja abstraktnih konceptov), učenje po iteracijah in vizualizacijo na večji množici podatkov. The abstractness of program concepts and the lack of students’ interest are the two main reasons for the high failure rate in introductory programming courses and the high dropout rate of students in colleges. To help and motivate students, researchers are constantly proposing various tools. One of many such methods is visualization of algorithms, which improves the quality of lectures and facilitates learning of algorithms. The reason is that the human brain processes visual images more easily than long pieces of source code, making it easier for a person to understand the process of implementing algorithms. The aim of this diploma thesis is to find different approaches and to lower the level of the abstraction of algorithms to the beginners, so the focus is on a group of students who are interested in the field of computer science and informatics and do not have prior knowledge of sorting algorithms or they just learn them.The application created in this diploma theses involves step-by-step learning (this is one of the most effective ways to understand abstract concepts), learning by iteration, and visualization on large datasets.

V svetu programiranja je urejanje podatkov zelo pomembno. Sčasoma so ljudje implementirali veliko različnih algoritmov, ki rešujejo ta problem ali pa izboljšujejo že obstoječe. Diplomsko delo obravnava problem urejanja podatkov na podlagi manj znanih algoritmov. Implementirani in analizirani so naslednji: sodo-liho urejanje (Odd-even sort), comb sort, ciklično urejanje (Cycle sort), pigeonhole sort, urejanje z drevesom (Tree sort) in timsort. Primerjava algoritmov je izvedena na podlagi prostorske in časovne zahtevnosti, dejanskega časa izvajanja, števila primerjav, števila prireditev in števila rekurzivnih klicev (če jih imajo). Da bi bila analiza realna, so vsi algoritmi analizirani na podlagi podobnih testnih podatkov, ki zajemajo večino možnih scenarijev pri urejanju. In the world of programming, sorting is very important. Over time, people have implemented many different algorithms that solve this problem or improve existing ones. The diploma thesis deals with the problem of data sorting using less known algorithms. Some of them have been implemented and analyzed: Odd-even Sort, Comb sort, Cycle sort, Pigeonhole sort, Tree sort, and Tim sort. The comparison of algorithms is based on the time and space complexity, the actual execution time, the number of comparisons, the number of assignments, and the number of recursive calls (if any). To make the analysis realistic, all algorithms are analyzed on similar test data. The test data cover most of the possible scenarios that could occur during sorting.

Pri izdelavi vizualnega gradiva je pomembno, da so napisi na slikah in videu berljivi ne glede na ozadje. S tem problemom se srečujemo v številnih dejavnostih, še zlasti v oglaševalski in filmski industriji. Cilj diplomske naloge je razviti in preizkusiti metodo določanja berljivosti napisov na poljubnih ozadjih. Učno množico smo oblikovali s pomočjo anketiranja, in sicer tako, da smo za večjo množico slik anketirance povprašali, ali so berljive. Nato smo zajeli ključne podatke (kontrast, svetloba ipd.). Pri tem smo uporabili barvna modela RGB in HSL. Na osnovi zajetih podatkov smo zgradili linearni model. Odgovor na vprašanje berljivosti smo v okviru naloge obravnavali kot dvojiški, zato smo model zgradili s pomočjo logistične regresije. Zgrajeni model smo ovrednotili z metodami, kot sta AUC in prečno preverjanje. Končni klasifikacijski model je bil pri napovedovanju berljivosti napisov natančen v 68 odstotkih. Na podlagi rezultatov bo lahko oglaševalec med samodejno generiranimi oglasi izbral najbolj berljive. In visual media production (e.g., in marketing and film industry), it is important that the text on images and video is legible regardless of the background. The goal of the thesis is to develop and evaluate a method to determine the legibility of text on arbitrary backgrounds. The dataset was created using surveys. For a large dataset of photos, we asked the participants whether they are legible or not. Subsequently, we gathered key features (contrast, lightness etc.) by using the RGB and HSL color models. The gathered data were employed to build a linear model. Because we perceive legibility as binary, we used logistic regression. The model was evaluated using such methods as AUC and cross validation. The final classification model is 68% accurate at predicting legibility. Based on these results, advertisers can, from a set of generated ads, select the most legible.