Pregled bibliografske jedinice broj: 1210688
YOLOv3 algorithm for object detection in aerial photos using standard pre trained networks
YOLOv3 algorithm for object detection in aerial photos using standard pre trained networks // 4th International Conference on Smart and Sustainable Technologies (SpliTech) / Perković, Toni ; Vukojević, Katarina ; Rodrigues, Joel ; Nižetić, Sandro ; Patrono, Luigi ; Šolić, Petar (ur.).
Piscataway (NJ): Institute of Electrical and Electronics Engineers (IEEE), 2019. 1014746, 636 (poster, recenziran, sažetak, ostalo)
CROSBI ID: 1210688 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
YOLOv3 algorithm for object detection in aerial photos using standard pre trained networks
Autori
Ana Šarić Gudelj, Tea Marasović, Vladan Papić
Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, ostalo
ISBN
978-953-290-091-0
Skup
4th International Conference on Smart and Sustainable Technologies (SpliTech)
Mjesto i datum
Bol, Hrvatska; Split, Hrvatska, 18.06.2019. - 21.06.2019
Vrsta sudjelovanja
Poster
Vrsta recenzije
Recenziran
Ključne riječi
convolutional neural networks, objects detections, CUDA, YOLOv3, aerial photos
Sažetak
In this article we made test of object detection on three sets of images using YOLO v3 algorithm and deep neural network (DNN) Darknet framework. The concept of DNN Darknet framework and CUDA is explained, with accent on processing speed. Behavior of the used architecture with pre trained Alexnet and ImageNet convolutional neural network when aerial orthogonal photos are used as input was investigated. Obtained results were compared to results of standard images.
Izvorni jezik
Engleski
Znanstvena područja
Računarstvo
