黑料不打烊

Bakalauro baigiamajame darbe kuriamas autonomi拧kas grup臈s navigacinis tinklas leid啪iantis lokaliai koordinuoti bei steb臈ti grup臋 lokaliai per 寞renginio LCD ekran膮 esant blokuojamiems mobiliojo ry拧io ir GNSS signalams, nes sistema naudoja alternatyvias komunikacijos technologijas. Darbas yra kompleksinis, sudarytas i拧 dviej懦 dali懦: komunikacijos, sudaroma i拧 naudotojo 寞renginio komunikacijos moduli懦, ir navigacijos, kuri yra sudaryta i拧 naudotojo 寞renginio navigacijos moduli懦 ir bazini懦 navigacini懦 sto膷i懦. Komunikacijos dalis leid啪ia saugiai komunikuoti tinklo naudotojams naudojant LoRa technologij膮 ir AES 拧ifruot臋, o navigacijos dalis leid啪ia tinklo naudotojui nustatyti savo viet膮 nepriklausomai nuo GNSS ry拧io, nes naudojama kuriama pilnai ant啪emin臈 vietos nustatymo sistema. Komunikacijos dalies sistemos projektavime naudojamas efektyvus STM32U5 mikrovaldiklis u啪tikrinantis kuo ilgesn寞 寞renginio veikim膮 i拧 baterijos bei LoRa modulis, kuris leid啪ia komunikuoti 寞renginiams tarpusavyje dideliu atstumu. Navigacijos dalies sistema projektuojama modulin臈, kurios valdymui naudojamas na拧us STM32H7R3 mikrovaldiklis, su greitai pakei膷iamais radijo bang懦 signalo apdorojimo elementais, kurie u啪tikrina fiksuot膮 radijo bang懦 signalo v臈linim膮, kuris yra reikalingas siekiant nustatyti buvimo viet膮. Bakalauro darb膮 sudaro sistem懦 strukt奴r懦 ir funkcij懦 apra拧ymas, principini懦 elektrini懦 schem懦 ir spausdintini懦 plok拧膷i懦 projektavimas, maket懦 ir programinio kodo 寞gyvendinimas bei testavimas. Darbo apimtis - 75 p. teksto be pried懦, 48 pav., 18 literat奴ros 拧altini懦, 6 priedai.

The bachelor鈥檚 thesis presents the development of an autonomous group navigation network that enables local coordination and monitoring of team members via LCD screen even when mobile network and GNSS signals are unavailable thanks to the use of alternative communication technologies. The work is divided into two interrelated parts: communication, which is made of network user's communication modules, and navigation, which is made of network user's navigation modules and base navigation stations. In the communication part, secure messaging among network users is achieved through LoRa technology combined with AES encryption. An energy efficient STM32U5 microcontroller maximizes battery life, while the LoRa module provides long range connectivity between user devices. The navigation part allows each user to determine their position independently of GNSS reception by means of a fully terrestrial positioning system under development. This subsystem features a modular design and is managed by a high performance STM32H7R3 microcontroller, together with quickly replaceable radio-signal processing elements that have a fixed signal delay necessary for accurate position estimation. The thesis contains descriptions of systems architectures and functional requirements, the design of schematic diagrams and printed circuit boards, and the implementation and testing of hardware prototypes alongside the corresponding software code. In total, the document spans 75 pages of text excluding appendices, includes 48 figures,18 bibliographic sources and 6 appendices.

艩io baigiamojo darbo tikslas buvo sukurti, 寞gyvendinti ir i拧bandyti sistem膮 啪moni懦 reakcijos laikui matuoti 寞 skirtingus dirgiklius 鈥� vaizdinius, garsinius ir jutiminius. Sistema paremta STM32 mikrovaldikliu, kuris kontroliuoja dirgikli懦 generavim膮, reakcijos fiksavim膮 bei rezultat懦 analiz臋.

The aim of this bachelor thesis was to design, implement, and test a system for measuring human reaction time to different stimuli 鈥� visual, auditory, and tactile. The system is based on an STM32 microcontroller, which manages stimulus generation, reaction detection, and result analysis.

艩iame bakalauro baigiamajame darbe apra拧ytas ir 寞gyvendintas funkcionalus i拧manusis elektroninio pa拧to asistento 寞skiepis 鈥濵icrosoft Outlook鈥� sistemai, kurio tikslas 鈥� reik拧mingai suma啪inti elektronini懦 lai拧k懦 tvarkymui skiriam膮 laik膮 ir padidinti vartotojo darbo efektyvum膮. To siekiama automatizuojant esminius lai拧k懦 apdorojimo procesus, tokius kaip turinio analiz臈, automatinis r奴拧iavimas ir pagalba kuriant tekst膮. 艩ios pagrindin臈s funkcijos buvo realizuotos taikant 拧iuolaikinius programin臈s 寞rangos k奴rimo metodus bei debesijos technologijas, u啪tikrinant skland啪i膮 integracij膮 ir patogi膮 vartotojo patirt寞 tiesiogiai 鈥濷utlook鈥� aplinkoje.

This bachelor's thesis describes and implements a functional smart email assistant add-in for the Microsoft Outlook system, the purpose of which is to significantly reduce the time spent managing emails and increase user work efficiency. This is achieved by automating essential email processing tasks, such as content analysis, automatic sorting, and assistance in text creation. These core features were realized using modern software development methods and cloud technologies, ensuring smooth integration and a user-friendly experience directly within the Outlook environment.

Bakalauro baigiamajame darbe kuriamas autonomi拧kas grup臈s navigacinis tinklas leid啪iantis lokaliai koordinuoti bei steb臈ti grup臋 lokaliai per 寞renginio LCD ekran膮 esant blokuojamiems mobiliojo ry拧io ir GNSS signalams, nes sistema naudoja alternatyvias komunikacijos technologijas. Darbas yra kompleksinis, sudarytas i拧 dviej懦 dali懦: komunikacijos, sudaroma i拧 naudotojo 寞renginio komunikacijos moduli懦, ir navigacijos, kuri yra sudaryta i拧 naudotojo 寞renginio navigacijos moduli懦 ir bazini懦 navigacini懦 sto膷i懦. Komunikacijos dalis leid啪ia saugiai komunikuoti tinklo naudotojams naudojant LoRa technologij膮 ir AES 拧ifruot臋, o navigacijos dalis leid啪ia tinklo naudotojui nustatyti savo viet膮 nepriklausomai nuo GNSS ry拧io, nes naudojama kuriama pilnai ant啪emin臈 vietos nustatymo sistema. Komunikacijos dalies sistemos projektavime naudojamas efektyvus STM32U5 mikrovaldiklis u啪tikrinantis kuo ilgesn寞 寞renginio veikim膮 i拧 baterijos bei LoRa modulis, kuris leid啪ia komunikuoti 寞renginiams tarpusavyje dideliu atstumu. Navigacijos dalies sistema projektuojama modulin臈, kurios valdymui naudojamas na拧us STM32H7R3 mikrovaldiklis, su greitai pakei膷iamais radijo bang懦 signalo apdorojimo elementais, kurie u啪tikrina fiksuot膮 radijo bang懦 signalo v臈linim膮, kuris yra reikalingas siekiant nustatyti buvimo viet膮. Bakalauro darb膮 sudaro sistem懦 strukt奴r懦 ir funkcij懦 apra拧ymas, principini懦 elektrini懦 schem懦 ir spausdintini懦 plok拧膷i懦 projektavimas, maket懦 ir programinio kodo 寞gyvendinimas bei testavimas. Darbo apimtis - 75 p. teksto be pried懦, 48 pav., 18 literat奴ros 拧altini懦, 6 priedai.

The bachelor鈥檚 thesis presents the development of an autonomous group navigation network that enables local coordination and monitoring of team members via LCD screen even when mobile network and GNSS signals are unavailable thanks to the use of alternative communication technologies. The work is divided into two interrelated parts: communication, which is made of network user's communication modules, and navigation, which is made of network user's navigation modules and base navigation stations. In the communication part, secure messaging among network users is achieved through LoRa technology combined with AES encryption. An energy efficient STM32U5 microcontroller maximizes battery life, while the LoRa module provides long range connectivity between user devices. The navigation part allows each user to determine their position independently of GNSS reception by means of a fully terrestrial positioning system under development. This subsystem features a modular design and is managed by a high performance STM32H7R3 microcontroller, together with quickly replaceable radio-signal processing elements that have a fixed signal delay necessary for accurate position estimation. The thesis contains descriptions of systems architectures and functional requirements, the design of schematic diagrams and printed circuit boards, and the implementation and testing of hardware prototypes alongside the corresponding software code. In total, the document spans 75 pages of text excluding appendices, includes 48 figures,18 bibliographic sources and 6 appendices.

Signal懦 filtr懦 maketas. Bakalauro baigiamasis darbas elektronikos in啪inerijos laipsniui. Vilniaus Gedimino technikos universitetas. Vilnius, 2025, 50 p., 15 iliustr., 5 bibl., 7 priedai. Bakalauro baigiamojo darbo tikslas yra sukurti aktyvi懦j懦 signal懦 filtr懦 maket膮, skirt膮 edukaciniams tikslams. Maketo sistem膮 sudaro aparatin臈 dalis su pasirenkamais filtrais ir integruotu signalo generatoriumi, Raspberry Pi mikrokompiuteris ir asmeninis kompiuteris. Mikrokompiuteris per analoginis-skaitmenins keitikl寞 tur臈t懦 priimti elektrin寞 signal膮 i拧 filtr懦 maketo i拧臈jimo ir perduoti j寞 寞 kompiuter寞 naudojant TCP/IP ry拧寞 per 鈥濫thernet鈥� s膮saj膮. Duomenys yra atvaizduojami naudojant vartotojo s膮saj膮, kurt膮 naudojant MATLAB庐 programin臋 寞rang膮. Naudojant vartotojo s膮saj膮, vartotojas gali pasirinkti signalo form膮, da啪n寞, filtro tip膮 ir steb臈ti filtr懦 da啪nines amplitud臈s charakteristikas bei signal懦 laiko diagramas. Signal懦 filtr懦 maket膮 sudaro pasirenkami 啪em懦 ir auk拧t懦 da啪ni懦, ir juostinis aktyvieji filtrai, realizuoti Sallen-Key topologija.

Signal Filters Prototype. Bachelor's Final Thesis for an Electronics Engineering Degree. 黑料不打烊_独家黑料_吃瓜网51爆料. Vilnius, 2025, 50 p., 15 illustr., 5 bibl., 7 append. The aim of the Bachelor's final thesis is to create an active signal filters prototype for educational purposes. The prototype system consists of a hardware part with selectable filters and an integrated signal generator, a Raspberry Pi microcomputer, and a personal computer. The microcomputer, via an analog-to-digital converter, should receive an electrical signal from the filter prototype output and transmit it to the computer using TCP/IP communication via an Ethernet interface. Data is displayed using a user interface created using MATLAB庐 software. Using the user interface, the user can select the signal waveform, frequency, filter type, and observe the filters' frequency amplitude characteristics and signal time-domain diagrams. The signal filters prototype includes selectable low-pass, high-pass, and band-pass active filters, implemented with Sallen-Key topology.

Bakalauro darbo tikslas 鈥� sukurti i拧mani膮j膮 pagalbos sistem膮 akliesiems ir silpnaregiams vartotojams. Sistem膮 sudaro 鈥濬lask鈥� pagr寞stas serveris, operatoriaus s膮saja, sukurta naudojant 鈥濿indowsForms鈥�, ir mobilioji programa, sukurta naudojant 鈥濺eact Native鈥�. Programa leid啪ia vartotojams atpa啪inti objektus, garsiai skaityti tekst膮 ir identifikuoti valiut膮 naudojant i拧maniojo telefono kamer膮 ir teksto 寞 kalb膮 technologij膮. Vartotojai pasirenka norim膮 funkcij膮 ir padaro nuotrauk膮 naudodami savo i拧man懦j寞 telefon膮. Sistema siun膷ia vaizdo duomenis 寞 server寞, kur juos apdoroja apmokytas YOLO neuroninis tinklas. Operatoriaus s膮saja leid啪ia administratoriams steb臈ti ir tvarkyti surinktus duomenis, sekti objekt懦 atpa啪inimo tikslum膮 ir palaikyti nuolatin寞 AI modelio tobulinim膮. UML diagramos, duomen懦 baz臈s strukt奴ra ir vartotojo s膮sajos maketai buvo sukurti taip, kad palaikyt懦 sistemos k奴rim膮. I拧samus testavimas patvirtino sistemos patikimum膮, prieinamum膮 ir tinkamum膮 tikslinei vartotoj懦 grupei. Galutinis produktas suteikia esmin臋 pagalb膮, skatina nepriklausomyb臋 ir pagerina kasdien臋 akl懦j懦 naudotoj懦 patirt寞. Darb膮 sudaro 10 dali懦: 寞vadas, ap啪valga, sudarymas, projektavimas, 寞gyvendinimas, patikra, i拧vados, literat奴ros s膮ra拧as, priedai. Darbo apimtis 鈥� 64 p. teksto be pried懦, 44 iliustr., 3 lent., 33 拧altiniai.

The goal of this bachelor鈥檚 thesis is to develop a smart vision assistance system for blind and visually impaired users. The system includes a Flask-based server, an operator interface created using Windows Forms, and a mobile application built with React Native. The application allows users to recognize objects, read text aloud, and identify currency using a smartphone camera and text-to-speech technology. Users select a desired function and take a picture. The system sends image data to the server, where it is processed by a pre-trained YOLO neural network. The operator interface enables administrators to monitor and manage collected data, track object recognition accuracy, and support continuous improvements to the AI model. UML diagrams, database structure, and user interface mockups were designed to support the system鈥檚 development. Comprehensive testing confirmed the system鈥檚 reliability, accessibility, and usability for the target user group. The final product provides essential assistance, promoting independence and enhancing the daily experiences of blind users. The work consists of 10 parts: introduction, overview, compilation, design, implementation, verification, conclusions, references, appendices. The scope of the work is 64 p. of text without appendices, 44 illustrations, 3 tables, 33 sources.

Magistro darbe tiriama, kaip dirbtinio intelekto (DI) modeliai gali automati拧kai vertinti JIRA sistemoje registruojam懦 programin臈s 寞rangos problem懦 apra拧ymo kokyb臋 ir generuoti patikslinamuosius klausimus, siekiant pagreitinti problem懦 sprendimo proces膮. Tyrime taikyti skirtingi DI metodai: u啪klausos be papildomo mokymo (angl. zero-shot), konteksto paie拧ka semantin臈je vektori懦 erdv臈je (angl. retrieval-augmented generation, RAG), modelio mokymas su instrukcijomis (angl. supervised fine-tuning, SFT) bei preferencij懦 optimizavimas (angl. direct preference optimization, DPO), naudojant LoRA (angl. low-rank adaptation) adapterius. Vertinimas atliktas remiantis keturiomis kiekybin臈mis metrikomis (BLEU-1, BLEU-2, ROUGE-L, METEOR) ir kokybiniu vertinimu, pasitelkiant did寞j寞 kalbos model寞 kaip vertintoj膮 (angl. LLM-as-a-Judge). Nustatyta, kad papildomai apmokyti modeliai, naudojant LoRA, SFT ir DPO metod懦 kombinacij膮, generuoja techniniu po啪i奴riu tikslius, ai拧kius ir semanti拧kai pagr寞stus klausimus, kuri懦 kokyb臈 priart臈ja prie 啪mogaus formuluojam懦 u啪klaus懦 lygio. Tyrimo rezultatai pagrind啪ia si奴lomo sprendimo taikym膮 praktin臈je programin臈s 寞rangos palaikymo veikloje ir atveria galimybes tolesniam automatizavimui.

This master鈥檚 thesis investigates how artificial intelligence (AI) models can automatically evaluate the quality of firmware issue reports submitted in the JIRA system and generate clarification questions to accelerate the issue resolution process. The study explores several AI-based methods: zero-shot prompting, semantic context retrieval in vector space (retrieval-augmented generation, RAG), instruction-based model fine-tuning (supervised fine-tuning, SFT), and preference optimization (direct preference optimization, DPO), using low-rank adaptation (LoRA) adapters. Evaluation was conducted using four quantitative metrics (BLEU-1, BLEU-2, ROUGE-L, METEOR) and a qualitative approach in which a large language model acted as the evaluator (LLM-as-a-Judge). The results show that fine-tuned models using the combination of LoRA, SFT, and DPO generate technically accurate, clear, and semantically grounded questions that approach human-crafted quality. The findings support the applicability of the proposed solution in practical software maintenance and highlight further opportunities for automation.

Elektroninis sintezatorius. Bakalauro baigiamasis darbas informacini懦 sistem懦 in啪inerijos laipsniui. Vilniaus Gedimino technikos universitetas. Vilnius, 2025, 53 p., 20 iliustr., 2 lent., 22 bibl., 7 priedai, 1 脳A3 plak. Sukurta realiuoju laiku veikianti elektroninio sintezatoriaus sistema, valdoma tiek per MIDI klaviat奴r膮, tiek per internetin臋 s膮saj膮. Projekto tikslas 鈥� 寞gyvendinti stabil懦 garso generavimo sprendim膮, reaguojant寞 寞 naudotojo 寞vest寞 su minimaliu v臈linimu. Pasirinktas 鈥濺aspberry Pi鈥� kompiuteris, Flask serveris ir Python garso apdorojimo biblioteka. 漠gyvendinti ADSR kont奴ras, filtr懦 moduliai, s膮saj懦 sinchronizacija. Testavimo metu 寞rodyta, kad sistema veikia stabiliai, o visos u啪duotyje numatytos funkcijos realizuotos s臈kmingai.

Electronic synthesizer. Bachelor's thesis for the degree in Information systems Engineering. 黑料不打烊_独家黑料_吃瓜网51爆料. Vilnius, 2025, 53 pages, 20 illustrations, 2 tables, 22 references, 7 appendices, 1 x A3 poster. A real-time operating electronic synthesizer system was developed, controlled both via a MIDI keyboard and through a web interface. The objective of the project was to implement a stable sound generation solution that responds to user input with minimal delay. For this a Raspberry PI computer, Flask server and a Python audio processing library was used. An ADSR envelope, filter modules, and interface synchronization were implemented. Testing confirmed that the system operates reliably and that all functions defined in the task were successfully realized.

Bakalauro baigiamojo darbo pagrindin臈 u啪duotis 鈥� sukurti sistem膮, skirt膮 kelio dangos tipo, b奴kl臈s ir 拧iurk拧tumo segmentavimui vaizde. Pirmoje darbo dalyje atliekta analogi拧k懦 sistem懦 ap啪valg膮, i拧nagrin臈tos 寞vairios vaizdo segmentavimo architekt奴ros bei vie拧ai prieinami duomen懦 rinkiniai. Antroje darbo dalyje suformuota bendra sistemos strukt奴rin臈 schema, numatanti nuosekl懦 keturi懦 pagrindini懦 modeli懦 veikim膮. Tre膷ioje darbo dalyje sudarytas nuosavas duomen懦 rinkinys, apimantis 1387 originalias nuotraukas, darytas 寞vairiomis oro ir ap拧vietimo s膮lygomis. Ketvirtoje dalyje atliktas nuodugnus sukurtos sistemos vertinimas. Darbo rezultatas 鈥� sukurta ir pagr寞sta keturi懦 moduli懦 sistema, paremta optimizuotais YOLOv8n segmentavimo modeliais, gebanti tiksliai nustatyti kelio dangos pavir拧i懦, tip膮, b奴kl臋 ir 拧iurk拧tum膮, naudojant specialiai sukurt膮 1387 anotuot懦 vaizd懦 duomen懦 rinkin寞 bei pritaikytas duomen懦 apdorojimo, balansavimo ir modeli懦 optimizavimo strategijas. Visi keturi YOLOv8n architekt奴ros modeliai buvo individualiai apmokyti, kruop拧膷iai parenkant parametrus. Apmokyti modeliai buvo integruoti 寞 vientis膮 sistem膮.Sistema buvo s臈kmingai integruota ir i拧testuota tiek su i拧 anksto 寞ra拧ytais vaizdais, tiek realiuoju laiku, pasiek臈 daugiau nei 80鈥�% tikslum膮 pagrindin臈se metrikose ir 10鈥�12 KPS apdorojimo greit寞, taip atitinkant visus funkcinius ir informacinius reikalavimus. Darb膮 sudaro: 68 p. teksto, 寞vadas, penkios d臈stymo dalys, i拧vados, bibliografinis s膮ra拧as, santrauka ir priedai.

The main objective of the Bachelor's thesis is to develop a system for segmenting the type, condition, and roughness of road surfaces in images. The first part of the thesis presents a review of analogous systems, an analysis of various image segmentation architectures, and publicly available datasets. The second part outlines the overall structural diagram of the system, which defines the sequential operation of four main models. In the third part, a custom dataset was created, consisting of 1,387 original photos taken under various weather and lighting conditions. The fourth part provides a thorough evaluation of the developed system. The result of the thesis is a developed and validated four-module system based on optimized YOLOv8n segmentation models, capable of accurately identifying the surface, type, condition, and roughness of road pavement using a specially created dataset of 1,387 annotated images, along with tailored data processing, balancing, and model optimization strategies. All four YOLOv8n architecture models were individually trained with carefully selected parameters and integrated into a unified system. The system was successfully implemented and tested both with pre-recorded images and in real-time, achieving over 80% accuracy across key metrics and a processing speed of 10鈥�12 FPS, thus meeting all functional and informational requirements. The thesis consists of 68 pages of text, including an introduction, five main chapters, conclusions, a bibliography, a summary, and appendices.

艩io projekto tikslas yra sukurti i拧mani膮 apsaugos sistem膮, valdom膮 寞terptinio 鈥濺aspberry Pi 4B鈥� kompiuterio, kurio kompakti拧kumas leid啪ia diegti sprendim膮 ribotose erdv臈se. Siekdami u啪tikrinti skland懦 sistemos veikim膮 拧iame 寞renginyje, pasirinkome 鈥濸ython鈥� programavimo kalb膮 kartu su lengvu, ta膷iau patikimu 鈥濬lask鈥� karkasu. Kadruose esan膷i懦 objekt懦 aptikimui ir identifikavimui integravome 鈥瀁OLOv5 nano鈥� neuronin寞 tinkl膮 鈥� jis pasi啪ymi optimizuotu na拧umu ir paprasta diegimo proced奴ra, tod臈l puikiai tinka ribot懦 i拧tekli懦 aplinkai.

The aim of this project is to develop an intelligent security system controlled by a Raspberry Pi 4B embedded computer, whose compactness allows the solution to be deployed in confined spaces. In order to ensure the smooth operation of the system on this device, we have chosen the Python programming language in combination with the lightweight yet robust Flask framework. To detect and identify the objects in the frames, we integrated the YOLOv5 nano neural network - which has optimized performance and a simple deployment procedure, making it a perfect fit for resource-constrained environments.