1. Globoki modeli za detekcijo in prepoznavo obrazov v video vsebinah in slikahStefani Bojanić, 2025, magistrsko delo Opis: Človeški obraz predstavlja eno od najpomembnejših biometričnih značilnosti, saj združuje informacijo o identiteti, spolu, starosti in čustvenem izrazu. V tem okviru se detekcija in prepoznavanje obrazov kažeta kot dva neločljivo povezana procesa. V magistrski nalogi so predstavljeni ključni izzivi tega področja ter stanje razvoja, ki zajema vse od klasičnih metod do sodobnih pristopov z globokim učenjem, s poudarkom na konvolucijskih nevronskih mrežah. Razvoj in eksperimenti so bili izvedeni s programskim jezikom Python. V okolju Visual Studio Code smo tako razvili sistem za prepoznavanje obrazov z uporabo algoritma ArcFace.
Ključne besede: detekcija obrazov, prepoznavanje obrazov, globoko učenje, state of the art, konvolucijske nevronske mreže (CNN), ArcFace
Objavljeno v DKUM: 23.10.2025; Ogledov: 0; Prenosov: 8
 Celotno besedilo (4,66 MB) |
2. Controllable speech-driven gesture generation with selective activation of weakly supervised controlsKarlo Crnek, Matej Rojc, 2025, izvirni znanstveni članek Opis: Generating realistic and contextually appropriate gestures is crucial for creating engaging embodied conversational agents. Although speech is the primary input for gesture generation, adding controls like gesture velocity, hand height, and emotion is essential for generating more natural, human-like gestures. However, current approaches to controllable gesture generation often utilize a limited number of control parameters and lack the ability to activate/deactivate them selectively. Therefore, in this work, we propose the Cont-Gest model, a Transformer-based gesture generation model that enables selective control activation through masked training and a control fusion strategy. Furthermore, to better support the development of such models, we propose a novel evaluation-driven development (EDD) workflow, which combines several iterative tasks: automatic control signal extraction, control specification, visual (subjective) feedback, and objective evaluation. This workflow enables continuous monitoring of model performance and facilitates iterative refinement through feedback-driven development cycles. For objective evaluation, we are using the validated Kinetic–Hellinger distance, an objective metric that correlates strongly with the human perception of gesture quality. We evaluated multiple model configurations and control dynamics strategies within the proposed workflow. Experimental results show that Feature-wise Linear Modulation (FiLM) conditioning, combined with single-mask training and voice activity scaling, achieves the best balance between gesture quality and adherence to control inputs.
Ključne besede: gesture generation, objective evaluation, selective control activation, transformers, weakly supervised learning
Objavljeno v DKUM: 09.09.2025; Ogledov: 0; Prenosov: 4
Celotno besedilo (1,63 MB) |
3. |
4. Obdelava masivnih podatkov pametnih domov na oblačni infrastrukturi : magistrsko deloAleks Brenčič, 2025, magistrsko delo Opis: V magistrski nalogi predstavljamo načrtovanje in razvoj sistema za shranjevanje masovnih podatkov pametnih domov na oblačni infrastrukturi. V nalogi predstavljamo ključne tehnologije, naprave in orodja za zajemanje podatkov v pametnih domovih in shranjevanje in obdelovanje le-teh v oblačnem okolju. Za zajemanje in nadzor pametnega okolja smo uporabili platformo Home Assistant, za oblačno okolje pa Amazon Web Services. Podatki pametnih naprav se iz pametnega okolja pošiljajo na oblačno okolje preko MQTT-protokola, hranijo pa v podatkovni bazi DynamoDB. Oblačna infrastruktura je inicializirana z orodjem Terraform, in omogoča visoko skalabilnost. Cilj razvoja sistema je omogočanje shranjevanja masovnih podatkov pametnih domov z uporabo oblačno infrastrukture, s čimer želimo zagotoviti dolgoročno hrambo in varno obdelavo podatkov, obenem pa omogočiti lažjo integracijo z ostalimi storitvami v oblaku.
Ključne besede: pametni dom, Amazon Web Services, MQTT, oblačna platforma
Objavljeno v DKUM: 13.08.2025; Ogledov: 0; Prenosov: 17
Celotno besedilo (2,02 MB) |
5. |
6. Weakly-supervised multilingual medical NER for symptom extraction for low-resource languagesRigon Sallauka, Umut Arioz, Matej Rojc, Izidor Mlakar, 2025, izvirni znanstveni članek Opis: Patient-reported health data, especially patient-reported outcomes measures, are vital for improving clinical care but are often limited by memory bias, cognitive load, and inflexible questionnaires. Patients prefer conversational symptom reporting, highlighting the need for robust methods in symptom extraction and conversational intelligence. This study presents a weakly-supervised pipeline for training and evaluating medical Named Entity Recognition (NER) models across eight languages, with a focus on low-resource settings. A merged English medical corpus, annotated using the Stanza i2b2 model, was translated into German, Greek, Spanish, Italian, Portuguese, Polish, and Slovenian, preserving the entity annotations medical problems, diagnostic tests, and treatments. Data augmentation addressed the class imbalance, and the fine-tuned BERT-based models outperformed baselines consistently. The English model achieved the highest F1 score (80.07%), followed by German (78.70%), Spanish (77.61%), Portuguese (77.21%), Slovenian (75.72%), Italian (75.60%), Polish (75.56%), and Greek (69.10%). Compared to the existing baselines, our models demonstrated notable performance gains, particularly in English, Spanish, and Italian. This research underscores the feasibility and effectiveness of weakly-supervised multilingual approaches for medical entity extraction, contributing to improved information access in clinical narratives—especially in under-resourced languages.
Ključne besede: low-resource languages, machine translation, medical entity extraction, NER, NLP, patient-reported outcomes, weakly-supervised learning
Objavljeno v DKUM: 19.05.2025; Ogledov: 0; Prenosov: 4
Celotno besedilo (338,94 KB) |
7. Visokonatančen elektronski sistem generiranja in stabilizacije referenčnega signala : magistrsko deloJani Preskar, 2024, magistrsko delo Opis: Namen te naloge je razvoj in testiranje sistema za generiranje in stabilizacijo 10 MHz referenčnega signala, ki je ključnega pomena za natančno delovanje komunikacijskih sistemov. Uporabljene metode vključujejo napetostno krmiljen kristalni oscilator, modul globalnega pozicijskega sistema za sinhronizacijo ter algoritme za merjenje in prilagajanje frekvence. Rezultati kažejo, da naš modul omogoča bolj natančne meritve in boljšo stabilnost frekvence v primerjavi s prejšnjim sistemom. Med testiranjem smo odkrili nekaj programskih hroščev, ki smo jih odpravili. V prihodnosti bi lahko uporaba umetne inteligence dodatno izboljšala naš algoritem.
Ključne besede: referenčni signal, stabilizacija frekvence, generiranje signala, kristalni oscilator, merjenje frekvence
Objavljeno v DKUM: 06.02.2025; Ogledov: 0; Prenosov: 0 |
8. Sistem strojne opreme v zanki za razvoj kompleksnih sistemov : magistrsko deloAleksander Vrbek, 2025, magistrsko delo Opis: V magistrskem delu je opisan razvoj tiskanega vezja strojne opreme v zanki katera simulira elektromotor. Razvito tiskano vezje omogoča simulacijo faznih tokov, temperature elektromotorja in pozicije rotorja. Uporabniku omogoča izbiro različnih senzorjev za določanje pozicije rotorja, kot so Sin-Cos, Resolver, SSI in HALL. Komunikacija z računalnikom je omogočena prek povezave CAN.
Ključne besede: vgrajeni sistem, tiskano vezje, sistem strojne opreme v zanki, brezkrtačni motor
Objavljeno v DKUM: 06.02.2025; Ogledov: 0; Prenosov: 33
Celotno besedilo (4,13 MB) |
9. An end-to-end framework for extracting observable cues of depression from diary recordingsIzidor Mlakar, Umut Arioz, Urška Smrke, Nejc Plohl, Valentino Šafran, Matej Rojc, 2024, izvirni znanstveni članek Opis: Because of the prevalence of depression, its often-chronic course, relapse and associated disability, early detection and non-intrusive monitoring is a crucial tool for timely diagnosis and treatment, remission of depression and prevention of relapse. In this way, its impact on quality of life and well-being can be limited. Current attempts to use artificial intelligence for the early classification of depression are mostly data-driven and thus non-transparent and lack effective means to deal with uncertainties. Therefore, in this paper, we propose an end-to-end framework for extracting observable depression cues from diary recordings. Furthermore, we also explore its feasibility for automatic detection of depression symptoms using observable behavioural cues. The proposed end-to-end framework for extracting depression was used to evaluate 28 video recordings from the Symptom Media dataset and 27 recordings from the DAIC-WOZ dataset. We compared the presence of the extracted features between recordings of individuals with and without a depressive disorder. We identified several cues consistent with previous studies in terms of their differentiation between individuals with and without depressive disorder across both datasets among language (i.e., use of negatively valanced words, use of first-person singular pronouns, some features of language complexity, explicit mentions of treatment for depression), speech (i.e., monotonous speech, voiced speech and pauses, speaking rate, low articulation rate), and facial cues (i.e., rotational energy of head movements). The nature/context of the discourse, the impact of other disorders and physical/psychological stress, and the quality and resolution of the recordings all play an important role in matching the digital features to the relevant background. In this way, the work presented in this paper provides a novel approach to extracting a wide range of cues relevant to the classification of depression and opens up new opportunities for further research.
Ključne besede: digital biomarkers of depression, facial cues, speech cues, language cues, deep learning, end-to-end pipeline, artificial intelligence
Objavljeno v DKUM: 17.01.2025; Ogledov: 0; Prenosov: 13
Celotno besedilo (2,34 MB) |
10. Načrtovanje in verifikacija digitalnega bloka BiSS vmesnikaMatevž Ves, 2024, magistrsko delo Opis: Magistrsko delo obravnava izdelavo digitalnega BiSS (Angl. Bidirectional Serial Synchronous)
bloka, ki ga je mogoče integrirati v podrejeno napravo. BiSS protokol je bil razvit z namenom učinkovite
in zanesljive komunikacije na področju industrijske in senzorske komunikacije. Magistrsko delo prav
tako predstavlja postopek načrtovanja digitalnega dela integriranih vezij. Ta zajema izdelavo RTL opis,
sintezo, izdelavo položajnega načrta (Angl. Floorplan) ter verifikacijo.
Končni blok je testiran do frekvence 10MHz vhodnega taktnega signala na MA liniji, med drugim
pa zajema adaptivno časovno kontrolo (Angl. Ataptive Timeout), nastavljivo dolžino procesnih
podatkov, dostopa do registrov preko kontrolne komunikacije ter delovanje v verižni vezavi (Angl. Daisy
Chain).
Ključne besede: integrirano vezje, digitalno vezje, BiSS vmesnik
Objavljeno v DKUM: 22.10.2024; Ogledov: 0; Prenosov: 33
Celotno besedilo (5,65 MB) |
