Home         Authors   Papers   Year of conference   Themes   Organizations        To MES conference

Intel OpenVINO™ Toolkit: Performance Analysis of Generative Adversarial Neural Networks
Authors
	Zunin V.V.
	Romanov A.
Date of publication
	2021
DOI
	10.31114/2078-7707-2021-2-83-90
Abstract
	The paper provides an overview of the current state of the implementation of neural networks and their methods of execution. A detailed description of the main components of the Intel® OpenVINO ™ Toolkit for executing neural networks on various Intel hardware platforms (CPU, GPU, Neural Compute Stick 2) is considered. The works in which this toolkit is used are investigated and the conclusions made in them are presented. Before carrying out the experiment, the selection of various generative adversarial neural networks, data sets for testing them, and the process of launching neural networks on selected hardware platforms (four experimental stands with different hardware equipment) and obtaining results are described. Based on the results of the experiments, the obtained data are presented based on two metrics: performance in FPS and the cost of using the tested hardware platform to execute a neural network. Based on the data obtained, the analysis of the performance metrics of the previously selected generative adversarial neural networks is carried out without the use of tools, using the OpenVINO ™ Toolkit in the usual mode, as well as using the Multi-Device Plugin and Heterogeneous Plugin on several connected devices.
Keywords
	Intel, OpenVINO, GAN, neural networks.
Library reference
	Zunin V.V., Romanov A. Intel OpenVINO™ Toolkit: Performance Analysis of Generative Adversarial Neural Networks // Problems of Perspective Micro- and Nanoelectronic Systems Development - 2021. Issue 2. P. 83-90. doi:10.31114/2078-7707-2021-2-83-90
URL of paper
	http://www.mes-conference.ru/data/year2021/pdf/D019.pdf