Naslov
A comparative study of YOLOv5 models performance for image localization and classification

Autori
Horvat, Marko ; Jelečević, Ljudevit ; Gledec, Gordan

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Proceedings of the 33rd Central European Conference on Information and Intelligent System (CECIIS 2022) / Vrček, Neven ; Guàrdia, Lourdes ; Grd, Petra - Varaždin : Fakultet organizacije i informatike Sveučilišta u Zagrebu, 2022, 349-356

Skup
33rd Central European Conference on Information and Intelligent Systems (CECIIS 2022)

Mjesto i datum
Dubrovnik, Hrvatska, 21.09.2022. - 23.09.2022

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
computer vision, image classification, deep learning, deep convolutional neural networks, YOLO

Sažetak
YOLOv5 is one of the latest and often used versions of a very popular deep learning neural network used for various machine learning tasks, mainly in computer vision. The YOLO algorithm has steadily gained acceptance in the data science community due to its superior performance in complex and noisy data environments, availability, and ease of use in combination with widely used programming languages such as Python. This paper aims to compare different versions of the YOLOv5 model using an everyday image dataset and to provide researchers with precise suggestions for selecting the optimal model for a given problem type. The obtained results and the implemented YOLOv5 models are available for non-commercial use at: https://github.com/mhorvat/YOLOv5-models-comparison

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA