Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders
Convolutional Neural Network (CNN) has attained high accuracy and it has been widely employed in image recognition tasks. In recent times, deep learning-based modern applications are evolving and it poses a challenge in research and development of hardware implementation. Therefore, hardware optimiz...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2020-01-01
|
Series: | IEEE Open Journal of Circuits and Systems |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/9134378/ |
id |
doaj-4399c3d1392f42e6857ae5b3ac56c332 |
---|---|
record_format |
Article |
spelling |
doaj-4399c3d1392f42e6857ae5b3ac56c3322021-03-29T19:01:00ZengIEEEIEEE Open Journal of Circuits and Systems2644-12252020-01-011768710.1109/OJCAS.2020.30073349134378Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree AddersFasih Ud Din Farrukh0https://orcid.org/0000-0002-0178-8130Chun Zhang1https://orcid.org/0000-0001-9791-4500Yancao Jiang2https://orcid.org/0000-0003-0145-0209Zhonghan Zhang3https://orcid.org/0000-0002-1695-0604Ziqiang Wang4https://orcid.org/0000-0001-6567-0759Zhihua Wang5https://orcid.org/0000-0001-6567-0759Hanjun Jiang6https://orcid.org/0000-0003-4911-0748Institute of Microelectronics, Tsinghua University, Beijing, ChinaInstitute of Microelectronics, Tsinghua University, Beijing, ChinaInstitute of Microelectronics, Tsinghua University, Beijing, ChinaInstitute of Microelectronics, Tsinghua University, Beijing, ChinaInstitute of Microelectronics, Tsinghua University, Beijing, ChinaInstitute of Microelectronics, Tsinghua University, Beijing, ChinaInstitute of Microelectronics, Tsinghua University, Beijing, ChinaConvolutional Neural Network (CNN) has attained high accuracy and it has been widely employed in image recognition tasks. In recent times, deep learning-based modern applications are evolving and it poses a challenge in research and development of hardware implementation. Therefore, hardware optimization for efficient accelerator design of CNN remains a challenging task. A key component of the accelerator design is a processing element (PE) that implements the convolution operation. To reduce the amount of hardware resources and power consumption, this article provides a new processing element design as an alternate solution for hardware implementation. Modified BOOTH encoding (MBE) multiplier and WALLACE tree-based adders are proposed to replace bulky MAC units and typical adder tree respectively. The proposed CNN accelerator design is tested on Zynq-706 FPGA board which achieves a throughput of 87.03 GOP/s for Tiny-YOLO-v2 architecture. The proposed design allows to reduce hardware costs by 24.5% achieving a power efficiency of 61.64 GOP/s/W that outperforms the previous designs.https://ieeexplore.ieee.org/document/9134378/Convolutional neural networkbooth encoding multiplierWALLACE tree addersFPGAadder treeobject detection |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Fasih Ud Din Farrukh Chun Zhang Yancao Jiang Zhonghan Zhang Ziqiang Wang Zhihua Wang Hanjun Jiang |
spellingShingle |
Fasih Ud Din Farrukh Chun Zhang Yancao Jiang Zhonghan Zhang Ziqiang Wang Zhihua Wang Hanjun Jiang Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders IEEE Open Journal of Circuits and Systems Convolutional neural network booth encoding multiplier WALLACE tree adders FPGA adder tree object detection |
author_facet |
Fasih Ud Din Farrukh Chun Zhang Yancao Jiang Zhonghan Zhang Ziqiang Wang Zhihua Wang Hanjun Jiang |
author_sort |
Fasih Ud Din Farrukh |
title |
Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders |
title_short |
Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders |
title_full |
Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders |
title_fullStr |
Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders |
title_full_unstemmed |
Power Efficient Tiny Yolo CNN Using Reduced Hardware Resources Based on Booth Multiplier and WALLACE Tree Adders |
title_sort |
power efficient tiny yolo cnn using reduced hardware resources based on booth multiplier and wallace tree adders |
publisher |
IEEE |
series |
IEEE Open Journal of Circuits and Systems |
issn |
2644-1225 |
publishDate |
2020-01-01 |
description |
Convolutional Neural Network (CNN) has attained high accuracy and it has been widely employed in image recognition tasks. In recent times, deep learning-based modern applications are evolving and it poses a challenge in research and development of hardware implementation. Therefore, hardware optimization for efficient accelerator design of CNN remains a challenging task. A key component of the accelerator design is a processing element (PE) that implements the convolution operation. To reduce the amount of hardware resources and power consumption, this article provides a new processing element design as an alternate solution for hardware implementation. Modified BOOTH encoding (MBE) multiplier and WALLACE tree-based adders are proposed to replace bulky MAC units and typical adder tree respectively. The proposed CNN accelerator design is tested on Zynq-706 FPGA board which achieves a throughput of 87.03 GOP/s for Tiny-YOLO-v2 architecture. The proposed design allows to reduce hardware costs by 24.5% achieving a power efficiency of 61.64 GOP/s/W that outperforms the previous designs. |
topic |
Convolutional neural network booth encoding multiplier WALLACE tree adders FPGA adder tree object detection |
url |
https://ieeexplore.ieee.org/document/9134378/ |
work_keys_str_mv |
AT fasihuddinfarrukh powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders AT chunzhang powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders AT yancaojiang powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders AT zhonghanzhang powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders AT ziqiangwang powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders AT zhihuawang powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders AT hanjunjiang powerefficienttinyyolocnnusingreducedhardwareresourcesbasedonboothmultiplierandwallacetreeadders |
_version_ |
1724196031367741440 |