Iskanje po katalogu digitalne knjižnice

Opis: V diplomskem delu je prikazan razvoj aplikacije za pretvorbo slik formata RAW v format PNG, pri čemer uporabljamo dva pristopa. Prvi temelji na standardnih algoritmih za pretvorbo slik, drugi pa na strojnem učenju, kar pomeni, da je bil model naučen čim bolj natančno pretvarjati sliko iz enega formata v drugega. Ta aplikacija dobro služi fotografom, ki zajemajo slike v formatu RAW, saj jih lahko na svojem mobilnem telefonu pretvorijo kar na poti. Na koncu sta sledili primerjava in analiza rezultatov za ugotavljanje, kateri postopek prinaša boljše rezultate. Za učenje modela je bilo uporabljenih 16 različnih slik, algoritem pa je bil implementiran s pomočjo knjižnic. Pretvorba z algoritmom je poskrbela za kvalitetnejše slike, vendar je bila pretvorba z modelom včasih hitrejša. Pretvorjene slike so bile primerjane z metrikama PSNR in SSIM ter analizirane.
Ključne besede: raw, png, slika, pretvorba, strojno učenje
Objavljeno v DKUM: 08.05.2025; Ogledov: 0; Prenosov: 9
Celotno besedilo (2,93 MB)

Opis: Acts of fraud have become much more prevalent in the financial industry with the rise of technology and the continued economic growth in modern society. Fraudsters are evolving their approaches continuously to exploit the vulnerabilities of the current prevention measures in place, many of whom are targeting the financial sector. To overcome and investigate financial frauds, this paper presents STALITA, which is an innovative platform for the analysis of bank transactions. STALITA enables graph-based data analysis using a powerful Neo4j graph database and the Cypher query language. Additionally, a diversity of other supporting tools, such as support for heterogeneous data sources, force-based graph visualisation, pivot tables, and time charts, enable in-depth investigation of the available data. In the Results section, we present the usability of the platform through real-world case scenarios.
Ključne besede: Neo4j, platform, bank transactions, graph analysis, graph visualisation, fraud, investigation
Objavljeno v DKUM: 27.03.2025; Ogledov: 0; Prenosov: 2
Celotno besedilo (1,44 MB)
Gradivo ima več datotek! Več...

Opis: Point clouds have become a popular training data for many practical applications of machine learning in the fields of environmental modeling and precision agriculture. In order to reduce high space requirements and the effect of noise in the data, point clouds are often transformed to a structured representation such as a voxel grid. Storing, transmitting and consuming voxelized geometry, however, remains a challenging problem for machine learning pipelines running on devices with limited amount of on-chip memory with low access latency. A viable solution is to store the data in a compact encoded format, and perform on-the-fly decoding when it is needed for processing. Such on-demand expansion must be fast in order to avoid introducing substantial additional delay to the pipeline. This can be achieved by parallel decoding, which is particularly suitable for massively parallel architecture of GPUs on which the majority of machine learning is currently executed. In this paper, we present such method for efficient and parallelizable encoding/decoding of voxelized geometry. The method employs multi-level context-aware prediction of voxel occupancy based on the extracted binary feature prediction table, and encodes the residual grid with a pointerless sparse voxel octree (PSVO). We particularly focused on encoding the datasets of voxelized trees, obtained from both synthetic tree models and LiDAR point clouds of real trees. The method achieved 15.6% and 12.8% reduction of storage size with respect to plain PSVO on synthetic and real dataset, respectively. We also tested the method on a general set of diverse voxelized objects, where an average 11% improvement of storage space was achieved.
Ključne besede: voxel grid, feature prediction, tree models, prediction-based encoding, key voxels, residuals, sparse voxel octree
Objavljeno v DKUM: 09.01.2025; Ogledov: 0; Prenosov: 5
Celotno besedilo (20,93 MB)

Opis: The 2024 Student Computing Research Symposium (SCORES 2024), organized by the Faculty of Electrical Engineering and Computer Science at the University of Maribor (UM FERI) in collabora-tion with the University of Ljubljana and the University of Primorska, showcases innovative student research in computer science. This year’s symposium highlights advancements in fields such as ar-tificial intelligence, data science, machine learning algorithms, computational problem-solving, and healthcare data analysis. The primary goal of SCORES 2024 is to provide a platform for students to present their research, fostering early engagement in academic inquiry. Beyond research presen-tations, the symposium seeks to create an environment where students from different institutions can meet, exchange ideas, and build lasting connections. It aims to cultivate friendships and future research collaborations among emerging scholars. Additionally, the conference offers an opportu-nity for students to interact with senior researchers from institutions beyond their own, promoting mentorship and broader academic networking.
Ključne besede: evaluacija, optimizacija, strojno učenje, podatki, zborniki
Objavljeno v DKUM: 26.11.2024; Ogledov: 0; Prenosov: 77
Celotno besedilo (35,41 MB)
Gradivo ima več datotek! Več...

Opis: Diplomsko delo analizira učinkovitost štirih formatov za stiskanje slik: PNG, WebP, JPEG2000 in HEIF. Glavni cilj dela je primerjava teh formatov na enakih slikah, da se oceni njihova kakovost po stiskanju, faktor stiskanja ter hitrost stiskanja in razširjanja. Po podrobnem opisu posameznih formatov, delo ponuja analizo in primerjavo rezultatov, pri čemer se HEIF izkaže za najučinkovitejšega v stiskanju, kljub počasnejšem času obdelave. WebP nudi najboljše ravnotežje med učinkovitostjo stiskanja in hitrostjo obdelave, medtem ko se JPEG2000 izkaže za dober kompromis med kakovostjo in faktorjem stiskanja, vendar je počasnejši pri stiskanju in razširjanju.
Ključne besede: PNG, WebP, JPEG2000, HEIF, stiskanje slik
Objavljeno v DKUM: 07.10.2024; Ogledov: 0; Prenosov: 21
Celotno besedilo (1,54 MB)