Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing

The problem existing in the transformation of deep learning algorithm to engineering application is analyzed. Combining with the characteristics and development trends of army intelligent ammunition, the missile-borne image processing heterogeneous accelerate system for deep learning is put forward...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autor principal:	Chen Dong, Tian Zonghao
Formato:	article
Lenguaje:	ZH
Publicado:	Editorial Office of Aero Weaponry 2021
Materias:	\|missile-borne image\|deep learning\|fpga\|systolic array\|winograd convolution Motor vehicles. Aeronautics. Astronautics TL1-4050
Acceso en línea:	https://doaj.org/article/992df31c0b834b0d8fa5910811648e78
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:992df31c0b834b0d8fa5910811648e78
record_format	dspace
spelling	oai:doaj.org-article:992df31c0b834b0d8fa5910811648e782021-11-30T00:13:41ZResearch on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing1673-504810.12132/ISSN.1673-5048.2020.0101https://doaj.org/article/992df31c0b834b0d8fa5910811648e782021-06-01T00:00:00Zhttps://www.aeroweaponry.avic.com/fileup/1673-5048/PDF/1628469357846-1777287579.pdfhttps://doaj.org/toc/1673-5048The problem existing in the transformation of deep learning algorithm to engineering application is analyzed. Combining with the characteristics and development trends of army intelligent ammunition, the missile-borne image processing heterogeneous accelerate system for deep learning is put forward based on the research of compression, quantitative and hardware heterogeneous acceleration, realizing heterogeneous hardware design. The DNNDK is used to compress and quantify the Yolo v3 model. The weight and parameter compression rate are more than 90% and 80%, realizing the lightweight design of Yolo v3. Based on the DPU hardware acceleration architecture, the algorithm is transplanted to the missile-borne embedded platform, and its power consumption and detection efficiency meet the requirements of missile-borne image processing.Chen Dong, Tian ZonghaoEditorial Office of Aero Weaponryarticle\|missile-borne image\|deep learning\|fpga\|systolic array\|winograd convolutionMotor vehicles. Aeronautics. AstronauticsTL1-4050ZHHangkong bingqi, Vol 28, Iss 3, Pp 10-17 (2021)
institution	DOAJ
collection	DOAJ
language	ZH
topic	\|missile-borne image\|deep learning\|fpga\|systolic array\|winograd convolution Motor vehicles. Aeronautics. Astronautics TL1-4050
spellingShingle	\|missile-borne image\|deep learning\|fpga\|systolic array\|winograd convolution Motor vehicles. Aeronautics. Astronautics TL1-4050 Chen Dong, Tian Zonghao Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing
description	The problem existing in the transformation of deep learning algorithm to engineering application is analyzed. Combining with the characteristics and development trends of army intelligent ammunition, the missile-borne image processing heterogeneous accelerate system for deep learning is put forward based on the research of compression, quantitative and hardware heterogeneous acceleration, realizing heterogeneous hardware design. The DNNDK is used to compress and quantify the Yolo v3 model. The weight and parameter compression rate are more than 90% and 80%, realizing the lightweight design of Yolo v3. Based on the DPU hardware acceleration architecture, the algorithm is transplanted to the missile-borne embedded platform, and its power consumption and detection efficiency meet the requirements of missile-borne image processing.
format	article
author	Chen Dong, Tian Zonghao
author_facet	Chen Dong, Tian Zonghao
author_sort	Chen Dong, Tian Zonghao
title	Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing
title_short	Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing
title_full	Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing
title_fullStr	Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing
title_full_unstemmed	Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing
title_sort	research on heterogeneous acceleration of deep learning method for missile-borne image processing
publisher	Editorial Office of Aero Weaponry
publishDate	2021
url	https://doaj.org/article/992df31c0b834b0d8fa5910811648e78
work_keys_str_mv	AT chendongtianzonghao researchonheterogeneousaccelerationofdeeplearningmethodformissileborneimageprocessing
_version_	1718406877543399424

Research on Heterogeneous Acceleration of Deep Learning Method for Missile-Borne Image Processing

Ejemplares similares