Pregled bibliografske jedinice broj: 1053063
Comparison of Semi-Global Block Matching Algorithm and DispNet Neural Network on KITTI Data Set
Comparison of Semi-Global Block Matching Algorithm and DispNet Neural Network on KITTI Data Set // 27th International Electrotechnical and Computer Science Conference (ERK 2018) / Žemva, Andrej ; Trost, Andrej (ur.).
Ljubljana: Društvo Slovenska sekcija IEEE, 2019. (predavanje, međunarodna recenzija, ostalo)
CROSBI ID: 1053063 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Comparison of Semi-Global Block Matching Algorithm and DispNet Neural Network on KITTI Data Set
Autori
Kramberger, Tin ; Potočnik, Božidar
Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, ostalo, ostalo
Skup
27th International Electrotechnical and Computer Science Conference (ERK 2018)
Mjesto i datum
Portorož, Slovenija, 17.09.2018. - 18.09.2018
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
Comparison, KITTI, DispNet, Neural Network, NN
Sažetak
Disparity estimation is a challenging task with numerous real-world applications. There are two main approaches to this problem: the classic Semi-global block matching algorithm and a new approach using a trained convolutional neural network to estimate the disparity. This paper shows the advancement of disparity estimation in terms of accuracy. The accuracy of disparity estimation by using the Semi-global block matching algorithm (SGM) and trained DispNet convolutional neural network are assessed and compared in this paper. Results on the KITTI test data set show better performance of the Disp-Net neural network in terms of accuracy compared to SGM. It could be said that neural networks are taking the primacy by disparity estimation.
Izvorni jezik
Engleski
Znanstvena područja
Računarstvo
