Software and Hardware Cooperative Acceleration Technology for CNN

To meet requirements of building intelligent avionics systems, and improve the intelligent combat level of manned/unmanned aerial vehicles, the software and hardware cooperative acceleration technology for CNN is designed and implemented to solve complex problems such as target recognition, auxiliar...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autor principal:	Li Xinyao, Liu Feiyang, Wen Pengcheng, Li Peng
Formato:	article
Lenguaje:	ZH
Publicado:	Editorial Office of Aero Weaponry 2021
Materias:	\|intelligent computing\|hardware acceleration\|target detection\|model compression\|fpga Motor vehicles. Aeronautics. Astronautics TL1-4050
Acceso en línea:	https://doaj.org/article/1cd2baeac61d4e898190fc12953bdecb
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:1cd2baeac61d4e898190fc12953bdecb
record_format	dspace
spelling	oai:doaj.org-article:1cd2baeac61d4e898190fc12953bdecb2021-11-30T00:13:41ZSoftware and Hardware Cooperative Acceleration Technology for CNN1673-504810.12132/ISSN.1673-5048.2020.0011https://doaj.org/article/1cd2baeac61d4e898190fc12953bdecb2021-06-01T00:00:00Zhttps://www.aeroweaponry.avic.com/fileup/1673-5048/PDF/2020-00011.pdfhttps://doaj.org/toc/1673-5048To meet requirements of building intelligent avionics systems, and improve the intelligent combat level of manned/unmanned aerial vehicles, the software and hardware cooperative acceleration technology for CNN is designed and implemented to solve complex problems such as target recognition, auxiliary decision-making, and autonomous planning. Aiming at solving the conflicts between the huge amount of parameters and the limited storage resources for embedded environment, the neural network model is optimized with model structure optimization and quantization of parameters. Aiming at solving the conflicts between complex floating-point operations and the shortage of computing resources, the convolution accelerating operator and the pooling accelerating operator are designed based on Verilog HDL. The pipeline and full parallel method are used to achieve the purpose of acceleration. Through the synergy of software optimization and hardware accelerated, the inference process of convolutional neural network is accelerated. Two popular CNN networks, that are YOLOv3 and YOLOv3-Tiny, are used as examples to accelerate and verify on the Xilinx ZCU102 FPGA development board. The results show that compared with the original models, the parameters of the accelerated models can be compressed about 3/4. The inference speed of YOLOv3 is increased by nearly 65 times, and that of YOLOv3-Tiny is increased by about 23 times.Li Xinyao, Liu Feiyang, Wen Pengcheng, Li PengEditorial Office of Aero Weaponryarticle\|intelligent computing\|hardware acceleration\|target detection\|model compression\|fpgaMotor vehicles. Aeronautics. AstronauticsTL1-4050ZHHangkong bingqi, Vol 28, Iss 3, Pp 99-104 (2021)
institution	DOAJ
collection	DOAJ
language	ZH
topic	\|intelligent computing\|hardware acceleration\|target detection\|model compression\|fpga Motor vehicles. Aeronautics. Astronautics TL1-4050
spellingShingle	\|intelligent computing\|hardware acceleration\|target detection\|model compression\|fpga Motor vehicles. Aeronautics. Astronautics TL1-4050 Li Xinyao, Liu Feiyang, Wen Pengcheng, Li Peng Software and Hardware Cooperative Acceleration Technology for CNN
description	To meet requirements of building intelligent avionics systems, and improve the intelligent combat level of manned/unmanned aerial vehicles, the software and hardware cooperative acceleration technology for CNN is designed and implemented to solve complex problems such as target recognition, auxiliary decision-making, and autonomous planning. Aiming at solving the conflicts between the huge amount of parameters and the limited storage resources for embedded environment, the neural network model is optimized with model structure optimization and quantization of parameters. Aiming at solving the conflicts between complex floating-point operations and the shortage of computing resources, the convolution accelerating operator and the pooling accelerating operator are designed based on Verilog HDL. The pipeline and full parallel method are used to achieve the purpose of acceleration. Through the synergy of software optimization and hardware accelerated, the inference process of convolutional neural network is accelerated. Two popular CNN networks, that are YOLOv3 and YOLOv3-Tiny, are used as examples to accelerate and verify on the Xilinx ZCU102 FPGA development board. The results show that compared with the original models, the parameters of the accelerated models can be compressed about 3/4. The inference speed of YOLOv3 is increased by nearly 65 times, and that of YOLOv3-Tiny is increased by about 23 times.
format	article
author	Li Xinyao, Liu Feiyang, Wen Pengcheng, Li Peng
author_facet	Li Xinyao, Liu Feiyang, Wen Pengcheng, Li Peng
author_sort	Li Xinyao, Liu Feiyang, Wen Pengcheng, Li Peng
title	Software and Hardware Cooperative Acceleration Technology for CNN
title_short	Software and Hardware Cooperative Acceleration Technology for CNN
title_full	Software and Hardware Cooperative Acceleration Technology for CNN
title_fullStr	Software and Hardware Cooperative Acceleration Technology for CNN
title_full_unstemmed	Software and Hardware Cooperative Acceleration Technology for CNN
title_sort	software and hardware cooperative acceleration technology for cnn
publisher	Editorial Office of Aero Weaponry
publishDate	2021
url	https://doaj.org/article/1cd2baeac61d4e898190fc12953bdecb
work_keys_str_mv	AT lixinyaoliufeiyangwenpengchenglipeng softwareandhardwarecooperativeaccelerationtechnologyforcnn
_version_	1718406849425833984

Software and Hardware Cooperative Acceleration Technology for CNN

Ejemplares similares