Open AccessResearch and Design of Open Convolutional Neural Network Based on FPGA
Author(s) -
Huining Lin,
Yu Wang,
Jian Zhou,
Shengqing Pei,
Zhiqiang Geng,
Chen Zhuang,
Zhihan Guo
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/714/4/042015
Subject(s) - field programmable gate array , convolutional neural network , computer science , benchmark (surveying) , scheduling (production processes) , computer architecture , embedded system , open set , artificial intelligence , pattern recognition (psychology) , engineering , geodesy , discrete mathematics , geography , operations management , mathematics
In this paper, using the fluidity and parallelism of FPGA, we design a set of memory scheduling mechanism to implement an open convolutional neural network. The self-built vehicle database and the cropped and zoomed Daimler Pedestrian Detection Benchmark dataset are each trained under different network structures, and their recognition rates reached 97.8% and 98.6%. An open feedforward network is implemented on the FPGA platform, and experiments show that for small convolutional networks with different structures, the FPGA platform can increase its recognition speed when resources permit.