[1]陈 禹,谷文成,渠吉庆,等.基于 PYNQ 的图像分类识别技术研究与实现[J].计算机技术与发展,2021,31(12):73-77.[doi:10. 3969 / j. issn. 1673-629X. 2021. 12. 013]
　CHEN Yu,GU Wen-cheng,QU Ji-qing,et al.Research and Implementation of Image Classification and Recognition Technology Based on PYNQ[J].,2021,31(12):73-77.[doi:10. 3969 / j. issn. 1673-629X. 2021. 12. 013]
点击复制

基于 PYNQ 的图像分类识别技术研究与实现()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
31
期数:
2021年12期
页码:
73-77
栏目:
图形与图像
出版日期:
2021-12-10

文章信息/Info

Title:
Research and Implementation of Image Classification and Recognition Technology Based on PYNQ
文章编号:
1673-629X(2021)12-0073-05
作者:
陈 禹1 ; 谷文成1 ; 渠吉庆1 ; 蒋志鹏1 ; 张 瑛1 ; 孙科学1; 2*
1. 南京邮电大学 电子与光学工程学院、微电子学院,江苏 南京 210023;
2. 射频集成与微组装技术国家地方联合工程实验室,江苏 南京 210023
Author(s):
CHEN Yu1 ; GU Wen-cheng1 ; QU Ji-qing1 ; JIANG Zhi-peng1 ; ZHANG Ying1 ; SUN Ke-xue1; 2 *
1. School of Electronic and Optical Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;
2. Nation-Local Joint Project Engineering Lab of RF Integration & Micropackage,Nanjing 210023,China
关键词:
卷积神经网络; 软硬件协同设计; PYNQ; VIVADO; Jupyter Notebook
Keywords:
convolutional neural network; software and hardware co-design; PYNQ; VIVADO; Jupyter Notebook
分类号:
TP391.4
DOI:
10. 3969 / j. issn. 1673-629X. 2021. 12. 013
摘要:
为了实现低功耗的图像分类识别系统,设计一种基于卷积神经网络的图像分类识别系统方案,该方案研究基于ARM+FPGA 异构系统的实现方法,系统搭载于 Xilinx 的 PYNQ 嵌入式开发平台。 在电脑端对待测试的数据集搭建卷积神经网络模型并完成 MNIST 和 CIFAR-10 数据集的训练验证。 随后设计特征参数提取函数完成权重和偏执参数的提取及格式转换,转换为硬件平台可以进行读取的二进制格式。 接着使用 Xilinx VIVADO HLS 设计工具,设计实现图像分类识别系统中卷积神经网络的自定义 IP 核模块。 完成自定义 IP 核的设计之后,以 IP 核模块和 ZYNQ 模块为主实现整体系统的通路搭建,完成验证后在 Jupyter Notebook 中通过上位机程序调用控制。 最后,完成驱动程序及系统上位机的设计。 测试结果表明,系统对 MNIST 和 CIFAR-10 数据集的识别可以实现分类,系统功耗仅为 1. 54 W。 该系统具有通用性好、硬件功耗低等优点,可广泛应用于边缘计算环境中。
Abstract:
In order to realize an image classification and recognition system with low power consumption, we propose an image classification and recogni-tion system solution based on convolutional neural network,which studies the implementation method of heterogeneous system based on ARM+ FPGA,and the system is mounted on Xilinx PYNQ embedded development platform. After building the convolutional neural network model on? ?the computer-side for the tested data sets,the training and verification of MNIST and CIFAR-10 data sets are completed. After that,the feature parameter extraction function is designed to complete the weight and paranoid parameter extraction and format conversion,which is converted? ? to a binary format that can be read by the hardware platform. Then Xilinx VIVADOHLS design tool is to design and implement a custom IP core module for the convolutional neural network in the image classification and recognition system. After the design of the custom IP core is comple-ted, the IP core module and the ZYNQ module are mainly used to implement the path construction of the overall system. After verification, the upper computer program is used to call the control in Jupyter Note book. Finally,the design of the driver and the host computer is completed, and the system is tested for functions and performance.The test shows that MNIST and CIFAR-10 data sets can be recognized and classified by the system,and the power consumption of the system is only 1. 54 W. The image classification and recognition system has the advantages of versatility, energy-saving,etc. ,which can be widely used in edge computing environment.

相似文献/References:

[1]叶棽,谢捷如. 基于SoC和嵌入式Linux的数据采集系统设计[J].计算机技术与发展,2015,25(08):203.
　YE Shen,XIE Jie-ru. Design of Data Acquisition System Based on SoC and Embedded Linux[J].,2015,25(12):203.
[2]张丹丹,李雷. 基于PCANet-RF的人脸检测系统[J].计算机技术与发展,2016,26(02):31.
　ZHANG Dan-dan,LI Lei. Face Detection System Based on PCANet-RF[J].,2016,26(12):31.
[3]邓宗平,赵启军,陈虎. 基于深度学习的人脸姿态分类方法[J].计算机技术与发展,2016,26(07):11.
　DEND Zong-ping,ZHAO Qi-jun,CHEN Hu. Face Pose Classification Method Based on Deep Learning[J].,2016,26(12):11.
[4]戴晓薇,赵启军. 基于回归的指纹方向场估计[J].计算机技术与发展,2017,27(01):1.
　DAI Xiao-wei,ZHAO Qi-jun. Fingerprint Orientation Field Estimation Based on Regression[J].,2017,27(12):1.
[5]李宏林. 分析式纹理合成技术及其在深度学习的应用[J].计算机技术与发展,2017,27(11):7.
　LI Hong-lin. Analyzed Texture-synthesis Techniques and Their Applications in Deep Learning[J].,2017,27(12):7.
[6]周成伟. 基于卷积神经网络的自然场景中数字识别[J].计算机技术与发展,2017,27(11):101.
　ZHOU Cheng-wei. Recognition of Numbers in Natural Scene with Convolutional Neural Network[J].,2017,27(12):101.
[7]曲之琳,胡晓飞.基于改进激活函数的卷积神经网络研究[J].计算机技术与发展,2017,27(12):77.[doi:10.3969/ j. issn.1673-629X.2017.12.017]
　QU Zhi-lin,HU Xiao-fei.Research on Convolutional Neural Network Based on Improved Activation Function[J].,2017,27(12):77.[doi:10.3969/ j. issn.1673-629X.2017.12.017]
[8]傅 鹏,谢世朋.基于级联卷积神经网络的车牌定位[J].计算机技术与发展,2018,28(01):134.[doi:10.3969/ j. issn.1673-629X.2018.01.029]
　FU Peng,XIE Shi-peng.License Plate Location Based on Cascaded Convolution Neural Network[J].,2018,28(12):134.[doi:10.3969/ j. issn.1673-629X.2018.01.029]
[9]崔凤焦.表情识别算法研究进展与性能比较[J].计算机技术与发展,2018,28(02):145.[doi:10．3969/j．issn．1673－629X．2018．02．031]
　CUI Feng-jiao.Ｒesearch and Performance Comparison of Facial Expression Ｒecognition Algorithm[J].,2018,28(12):145.[doi:10．3969/j．issn．1673－629X．2018．02．031]
[10]陈强锐,谢世朋.基于深度学习的肺部肿瘤检测方法[J].计算机技术与发展,2018,28(04):201.[doi:10.3969/ j. issn.1673-629X.2018.04.043]
　CHEN Qiang-rui,XIE Shi-peng.Lung Cancer Detection Method Based on Deep Learning[J].,2018,28(12):201.[doi:10.3969/ j. issn.1673-629X.2018.04.043]
[11]谷文成,高谷九祥,凌卓毅,等.基于 PYNQ 的手写体数字识别系统设计实现[J].计算机技术与发展,2022,32(S2):31.[doi:10. 3969 / j. issn. 1673-629X. 2022. S2. 005]
　GU Wen-cheng,GAO Gu-jiu-xiang,LING Zhuo-yi,et al.Design and Implementation of Handwritten Digit Recognition System on Mobile Terminal Based on PYNQ[J].,2022,32(12):31.[doi:10. 3969 / j. issn. 1673-629X. 2022. S2. 005]

更新日期/Last Update: 2021-12-10