prikaz prve stranice dokumenta Implementacija neuronske mreže na ZYBO razvojnom sustavu
  Download
PDF 3.35 MB
master's thesis
Implementacija neuronske mreže na ZYBO razvojnom sustavu
2019. urn:nbn:hr:200:904565
Birka, Toni
Josip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
Department of Computer Engineering and Automation
Chair of Computer Engineering

Cite this document

Birka, T. (2019). Implementacija neuronske mreže na ZYBO razvojnom sustavu (Master's thesis). Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek. Retrieved from https://urn.nsk.hr/urn:nbn:hr:200:904565

Birka, Toni. "Implementacija neuronske mreže na ZYBO razvojnom sustavu." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. https://urn.nsk.hr/urn:nbn:hr:200:904565

Birka, Toni. "Implementacija neuronske mreže na ZYBO razvojnom sustavu." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. https://urn.nsk.hr/urn:nbn:hr:200:904565

Birka, T. (2019). 'Implementacija neuronske mreže na ZYBO razvojnom sustavu', Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, accessed 19 November 2024, https://urn.nsk.hr/urn:nbn:hr:200:904565

Birka T. Implementacija neuronske mreže na ZYBO razvojnom sustavu [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek; 2019 [cited 2024 November 19] Available at: https://urn.nsk.hr/urn:nbn:hr:200:904565

T. Birka, "Implementacija neuronske mreže na ZYBO razvojnom sustavu", Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Osijek, 2019. Available at: https://urn.nsk.hr/urn:nbn:hr:200:904565

Please login to the repository to save this object to your list.
Metadata
TitleImplementacija neuronske mreže na ZYBO razvojnom sustavu
Title (english)Neural Network Implementation on ZYBO Development Board
AuthorToni Birka
MentorTomislav Matić (mentor)
Committee memberTomislav Matić (predsjednik povjerenstva)
Committee memberIvan Aleksi (član povjerenstva)
Committee memberIvan Vidović (član povjerenstva)
GranterJosip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
(Department of Computer Engineering and Automation)
(Chair of Computer Engineering)
Osijek
Defense date and country2019-09-30, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Architecture of Computer Systems
Abstract
Cilj rada bila je implementacija neuronske mreže na ZYBO razvojnom sustavu, što je postignuto razvijanjem ZYNN sustava. Grafičko korisničko sučelje sustava, koje je razvijeno koristeći Kivy framework, omogućuje dizajniranje i treniranje novih Keras modela i učitavanje postojećih modela prije slanja parametara mreže na ZYBO platformu. Nakon slanja parametara, slijedi prepoznavanje znamenaka sa slike kamere, koja je prenesena u memoriju platforme HDMI kabelom. Na platformi se prvo izvodi proces predobrade slike, kojim se otklanja šum, lociraju nakupine piksela, odnosno potencijalne znamenke, koje se zatim skaliraju i prilagođuju tako da budu što sličnije onima iz MNIST baze podataka. Nakon toga se svaka nakupina piksela provlači kroz model neuronske mreže čiji su parametri prethodno poslani. Rezultati klasifikacije se šalju i prikazuju na korisničkom sučelju na računalu. Osim toga, na monitoru, koji je spojen preko VGA sučelja, se prikazuje obrađena slika s kamere na kojoj su uokvirene nakupine piksela koje su klasificirane kao znamenka, odnosno one kojima je vjerojatnost pripadanja nekoj klasi veća od 55 %. Korisnik zatim može klasificirati drugi skup znamenaka ili odabrati ili kreirati neki drugi Keras model za klasifikaciju.
Abstract (english)
The aim of the paper was implementation of a neural network on the ZYBO development system, which was achieved by developing the ZYNN system. The graphical user interface of the system, developed using the Kivy framework, provides designing and training new Keras models and loading existing models before sending the network parameters to the ZYBO platform. After sending the parameters, the digits are recognized from the camera image, which is transferred to the platform memory via an HDMI cable. The platform first preprocesses an image that eliminates noise, locates pixel clusters, or potential digits, which are then scaled and adjusted to be as similar as possible to those of the MNIST database. Subsequently, each potential digit is passed through a neural network model whose parameters have been previously sent. The classification results are sent and displayed on the computer user interface. In addition, the monitor, which is connected via the VGA interface, displays a processed image from a camera with framed pixel clusters that are classified as a digit, that is, those that have a probability belonging to a class of more than 55 %. The user can then classify another set of digits or select or create another Keras model for classification.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:200:904565
Study programmeTitle: Graduate University Study Programme in Computer Engineering
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
Created on2020-09-04 08:39:47