prikaz prve stranice dokumenta Prepoznavanje objekata na RISC-V arhitekturi računala
  Download
PDF 1.33 MB
master's thesis
Prepoznavanje objekata na RISC-V arhitekturi računala
2025. urn:nbn:hr:168:944339
Budimir-Bekan, Borna
University of Zagreb
Faculty of Electrical Engineering and Computing

Cite this document

Budimir Bekan, B. (2025). Prepoznavanje objekata na RISC-V arhitekturi računala (Master's thesis). Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing. Retrieved from https://urn.nsk.hr/urn:nbn:hr:168:944339

Budimir Bekan, Borna. "Prepoznavanje objekata na RISC-V arhitekturi računala." Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2025. https://urn.nsk.hr/urn:nbn:hr:168:944339

Budimir Bekan, Borna. "Prepoznavanje objekata na RISC-V arhitekturi računala." Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2025. https://urn.nsk.hr/urn:nbn:hr:168:944339

Budimir Bekan, B. (2025). 'Prepoznavanje objekata na RISC-V arhitekturi računala', Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, accessed 18 March 2025, https://urn.nsk.hr/urn:nbn:hr:168:944339

Budimir Bekan B. Prepoznavanje objekata na RISC-V arhitekturi računala [Master's thesis]. Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing; 2025 [cited 2025 March 18] Available at: https://urn.nsk.hr/urn:nbn:hr:168:944339

B. Budimir Bekan, "Prepoznavanje objekata na RISC-V arhitekturi računala", Master's thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, 2025. Available at: https://urn.nsk.hr/urn:nbn:hr:168:944339

Please login to the repository to save this object to your list.
Metadata
TitlePrepoznavanje objekata na RISC-V arhitekturi računala
Title (english)Object Detection on RISC-V Platform
AuthorBorna Budimir-Bekan
MentorVladimir Čeperić (mentor)
Committee memberVladimir Čeperić (predsjednik povjerenstva)
Committee memberZoran Kalafatić (član povjerenstva)
Committee memberMarko Koričić (član povjerenstva)
GranterUniversity of Zagreb
Faculty of Electrical Engineering and Computing
Zagreb
Defense date and country2025-02-25, Croatia
Abstract
Računalni vid i detekcija objekata postali su ključni segmenti modernih sustava umjetne inteligencije, pri čemu se sve više istražuju alternativne hardverske platforme koje omogućuju njihovu izvedbu. Ovaj rad analizira mogućnosti primjene RISC-V arhitekture u kontekstu računalnog vida, s posebnim naglaskom na izvedbu detekcije objekata na VisionFive 2 jednopločnom računalu. U prvom dijelu rada predstavljene su temeljne značajke RISC-V arhitekture te usporedba s popularnijim ARM i x86... More rješenjima. Također se istražuju tehničke specifikacije VisionFive 2 uređaja, njegova kompatibilnost s modernim softverskim bibliotekama te ograničenja koja proizlaze iz nedovršene GPU podrške. Eksperimentalni dio rada obuhvaća implementaciju i testiranje YOLO i MobileNet SSD modela na VisionFive 2. Analizirane su performanse u pogledu brzine inferencije na CPU-u te potencijalni dobitci uz buduću GPU akceleraciju. Rezultati pokazuju da iako VisionFive 2 može izvršavati osnovne zadatke računalnog vida, postoje značajna ograničenja u odnosu na konkurentske SBC uređaje poput Raspberry Pi-a, prvenstveno zbog nedostatka optimiziranih drivera i podrške za hardversku akceleraciju. Zaključni dio rada ističe ključne izazove i moguća poboljšanja za buduću implementaciju računalnog vida na RISC-V platformama. Iako se RISC-V pokazuje kao perspektivna alternativa zatvorenim arhitekturama, potrebno je daljnje usavršavanje softverskog ekosustava kako bi se postigla bolja integracija s modernim AI modelima. Less
Abstract (english)
Computer vision and object detection have become essential components of modern artificial intelligence systems, prompting increased exploration of alternative hardware platforms for their implementation. This paper examines the feasibility of utilizing RISC-V architecture for computer vision applications, with a particular focus on object detection performance on the VisionFive 2 single-board computer. The first part of the paper introduces the fundamental features of RISC-V... More architecture and compares it with more widely adopted ARM and x86 solutions. Additionally, the technical specifications of the VisionFive 2 device are analyzed, including its compatibility with modern software libraries and the limitations arising from incomplete GPU support. The experimental section involves the implementation and benchmarking of YOLO and MobileNet SSD models on VisionFive 2. The study evaluates inference performance using the CPU and explores the potential benefits of future GPU acceleration. The results indicate that while VisionFive 2 can handle basic computer vision tasks, it faces significant limitations compared to competing SBC devices like the Raspberry Pi, primarily due to the lack of optimized drivers and hardware acceleration support. The conclusion highlights key challenges and potential improvements for the future development of computer vision applications on RISC-V platforms. Although RISC-V presents a promising alternative to proprietary architectures, further enhancements in its software ecosystem are required to achieve better integration with modern AI models. Less
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:168:944339
Study programmeTitle: Computing
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsOpen access
Terms of useLicence In copyright icon
Public note
Created on2025-03-12 11:05:30
accessibility

closeAccessibilityrefresh

If you wish to permanently save changes, click on Save, if not - your settings will be reset when you restart the browser.