基于FPGA的卷积神经网络加速模块设计
梅志伟,王维东

Design of Convolutional Neural Network acceleration module based on FPGA
Zhiwei Mei,Weidong Wang
表1 常见卷积神经网络参数分析
Table 1 Parameteric analysis of common CNNs
卷积神经网络AlexNetVGG?16GoogleNet v1Yolo v3
卷积核尺寸3,5,1131,3,5,71,3,7
除第一层外通道数64~25632~51232~83232~1024
卷积层参数量2.3 M14.7 M6.0 M40.5 M
卷积层乘加数量666 M15.3 G1.43 G65.4 G
全连接层参数量58.6 M124 M1 M-
全连接层乘加数量58.6 M130 M1 M-