prikaz prve stranice dokumenta Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje
  Download
PDF 1.56 MB
master's thesis
Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje
2023. urn:nbn:hr:200:410517
Pocrnja, Josip
Josip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
Department of Computer Engineering and Automation
Chair of Automation and Robotics

Cite this document

Pocrnja, J. (2023). Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje (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:200:410517

Pocrnja, Josip. "Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2023. https://urn.nsk.hr/urn:nbn:hr:200:410517

Pocrnja, Josip. "Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2023. https://urn.nsk.hr/urn:nbn:hr:200:410517

Pocrnja, J. (2023). 'Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje', Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, accessed 02 May 2024, https://urn.nsk.hr/urn:nbn:hr:200:410517

Pocrnja J. Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek; 2023 [cited 2024 May 02] Available at: https://urn.nsk.hr/urn:nbn:hr:200:410517

J. Pocrnja, "Procjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje", Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Osijek, 2023. Available at: https://urn.nsk.hr/urn:nbn:hr:200:410517

Please login to the repository to save this object to your list.
Metadata
TitleProcjena dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje
Title (english)Depth estimation using monocular image in autonomous driving
AuthorJosip Pocrnja
MentorRatko Grbić (mentor)
Committee memberRatko Grbić (član povjerenstva)
Committee memberMario Vranješ (predsjednik povjerenstva)
Committee memberMarijan Herceg (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 Automation and Robotics)
Osijek
Defense date and country2023-10-05, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Artificial Intelligence
Abstract
U ovom radu opisane su postojeće metode procjene dubinske mape na temelju slike dobivene s jedne kamere u okviru autonomne vožnje. Objašnjena je korištena neuronska mreža PixelFormer koja je zasnovana na swin transformeru. Dobivena mreža trenirana je i testirana na prikupljenim skupovima podataka iz CARLA simulatora. Način prikupljanja novog skupa podataka u simulaciji izveden je po uzoru na KITTI skup podataka koristeći kamere postavljene na prednjem kraju vozila. Dobiveni rezultat je neuronska mreža koja radi procjenu dubinske mape u simulacijskim uvjetima. Nadalje, predložen je algoritam za automatsko kočenje u slučaju nailaska vozila na prepreku koji koristi procijenjenu dubinsku mapu na temelju RGB slike. Cjelokupni algoritam izrađen je u Python programskom okruženju s dodatkom PyTorch biblioteke. Predloženi algoritam evaluiran je na dva načina. Najprije je evaluirana efikasnost procjene dubinske mape iz RGB kamere montirane na vozilu u simulatoru, a zatim je evaluirano automatsko kočenje vozila mjerenjem udaljenosti na kojoj se vozilo zaustavilo na različitim mapama s obzirom na različite prepreke.
Abstract (english)
This paper describes existing methods for estimating depth maps based on images obtained from a single camera within the context of autonomous driving. It explains the use of the PixelFormer neural network, which is based on the Swin Transformer architecture. The obtained network was trained and tested on datasets collected from the CARLA simulator. The data collection approach in the simulation was inspired by the KITTI dataset, using cameras mounted at the front of the vehicle. The result is a neural network that can estimate depth maps in simulated conditions. Furthermore, an algorithm for automatic braking in case of encountering obstacles is proposed, which utilizes the estimated depth map based on the RGB image. The algorithm is implemented in a Python programming environment with the addition of the PyTorch library. The proposed algorithm was evaluated in two ways. First, the efficiency of estimating the depth map from an RGB camera mounted on the vehicle in the simulator was evaluated. Then, automatic braking of the vehicle was evaluated by measuring the distance at which the vehicle came to a stop in different scenarios with various obstacles.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:200:410517
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 on2023-10-13 08:28:40