• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学 ›› 2021, Vol. 43 ›› Issue (04): 628-633.

• 高性能计算 • 上一篇    下一篇

基于FPGA的高效可伸缩的MobileNet加速器实现

萧嘉乐,梁东宝,陈弟虎,粟涛   

  1. (中山大学电子与信息工程学院,广东 广州 510275)
  • 收稿日期:2020-04-05 修回日期:2020-06-28 接受日期:2021-04-25 出版日期:2021-04-25 发布日期:2021-04-21
  • 基金资助:
    广东省重大科技计划(2017B090909005,2019B010140002)

An efficient and scalable MobileNet accelerator based on FPGA

XIAO Jia-le,LIANG Dong-bao,CHEN Di-hu,SU Tao#br#

#br#
  

  1. (School of Electronics and Information Technology,Sun Yat-sen University,Guangzhou 510275,China)

  • Received:2020-04-05 Revised:2020-06-28 Accepted:2021-04-25 Online:2021-04-25 Published:2021-04-21

摘要: MobileNet网络是一种广泛应用于嵌入式领域的深度神经网络,为了解决其硬件实现效率低的问题,同时达到在不同硬件资源下具有一定可伸缩性,提出了基于FPGA的一款MobileNet网络加速器结构,针对网络的堆叠结构特性设计了三级流水的加速阵列,并实现了在0~4000乘法器开销下都达到70%以上的计算效率。最终在XILINX Zynq-7000 ZC706开发板上实现了MoblieNet网络加速器,在150 MHz工作频率下,可达到156 Gop/s的性能和61%的计算效率,计算效率高于其他MobileNet网络加速器的。


关键词: MobileNet网络, 卷积神经网络, 硬件加速器

Abstract: MobileNet network is a deep neural network mode widely used in the embedded field. In order to solve the problem of low hardware implementation efficiency and achieve certain scalability under different hardware resources, a MobileNet network accelerator structure based on FPGA is proposed. According to the stacking structure characteristics of the network, a three-level pipeline acceleration array is designed, and the computing efficiency is over 70% within 4000 multipliers. A 150 MHz fully working demo on XILINX Zynq-7000 ZC706  development board achieves 156 Gop/s performance and 61% calculation efficiency, which is higher than other MobileNet network accelerators.


Key words: MobileNet network, convolutional neural network, hardware accelerator