Search
Deep learning for brands object detection and recognition in images
View/
Date
2019Publisher
BestprintCitation
Deep learning for brands object detection and recognition in images / V. А. Golovko, E. V. Mikhno, A. А. Kroschenko, S. V. Bezobrazov. – Text : direct // PRIP'2019. Pattern Recognition and Information Processing = Распознавание образов и обработка информации : proceedings of the 14th international conference, Minsk, 21–23 may 2019 / Belarusian state university of informatics and radioelectronics. – Minsk : Bestprint, 2019. – ISBN 978-985-90509-3-0. – P. 155–158. – Bibliography: 14 titles.Abstract
In this paper, we investigate applying several well-known models to the task of brands detection in images.
We implemented comparison of most effective and widely used architectures as Faster R-CNN (ResNet50/101), SSD и YOLO. Received results confirm the effectiveness of applying Faster R-CNN to any sets of images. However, it is necessary to note the resource-intensiveness of this architecture and its unsuitability for solving problems, in which an important criterion of efficiency is the time for performing the analysis. The SSD and YOLO models do not offer advantages in the detection of small and medium-sized objects, but can be successfully used as part of mobile detection systems that are limited in their hardware capabilities. In addition, these neural network architectures perform processing faster than Faster R- CNN and can be considered as basic models for detecting and segmentation of objects in images and video in real time.
Collection
Это произведение доступно по лицензии Creative Commons «Attribution-NonCommercial» («Атрибуция-Некоммерчески») 4.0 Всемирная.