摘 要:目标检测作为计算机视觉技术的基础任务,在智慧医疗、智能交通等生活场景中应用广泛。深度学习具有高类别检测精度、高精度定位的优势,是当前目标检测的研究重点。由于卷积神经网络计算复杂度高、内存要求高,使用 CPU 实现的设计方案已经难以满足实际应用的需求。现场可编程逻辑门阵列(FPGA)具有可重构、高能效、低延迟的特点。研究围绕FPAG 硬件设计,选取了 YOLOv2 算法,并针对该算法设计了对应的硬件加速器,实现了基于 FPGA 的目标检测。
关键词:YOLO;FPGA;目标检测;深度学习
DOI:10.19850/j.cnki.2096-4706.2023.07.026
基金项目:浙江省教育厅一般项目(202146495)
中图分类号:TP391.4 文献标识码:A 文章编号:2096-4706(2023)07-0101-04
Design of Object Detection System and Accelerator Based on FPGA
WU Yuhao
(Jiaxing Vocational & Technical College, Jiaxing 314036, China)
Abstract: As a basic task of computer vision technology, object detection is widely used in smart medicine, intelligent transportation and other life scenes. Deep Learning nowadays becomes the research focus of object detection for its advantages of high precision in class detection and positioning. Due to the great computational complexity and memory requirements, the design scheme implemented by using the CPU has been difficult to meet the needs of practical application. FPGA has the characteristics of reconfigurability, high energy efficiency and low latency. The research focuses on the FPGA hardware design, selects the YOLOv2 algorithm, designs the corresponding hardware accelerator in terms of the algorithm, and realizes the object detection based on FPGA.
Keywords: YOLO; FPGA; object detection; Deep Learning
参考文献:
[1] FANG L,HANGJIANG H E,ZHOU G. Research overview of object detection methods [J].Computer Engineering and Applications,2018,54(13):11-18.
[2] 赵兴博,陶青川 . 适用于 FPGA 的轻量实时视频人脸检测[J]. 现代计算机,2022,28(8):1-8.
[3] REDMON J,DIVVALA S,GIRSHICK R,et al. You Only Look Once:Unified,Real-Time Object Detection [C]//IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE,2016:779-778.
[4] 周飞燕,金林鹏,董军 . 卷积神经网络研究综述 [J]. 计算机学报,2017,40(6):1229-1251.
[5] 张珂,冯晓晗,郭玉荣,等 . 图像分类的深度卷积神经网络模型综述 [J]. 中国图象图形学报,2021,26(10):2305-2325.
[6] 吴艳霞,梁楷,刘颖,等 . 深度学习 FPGA 加速器的进展与趋势 [J]. 计算机学报,2019,42(11):2461-2480.
[7] 刘腾达,朱君文,张一闻 .FPGA 加速深度学习综述 [J].计算机科学与探索,2021,15(11):2093-2104.
[8] ZHANG C,LI P,SUN G,et al. Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks [C]//the 2015 ACM/SIGDA International Symposium. Monterey:ACM,2015.
[9] 裴颂文,汪显荣 .YOLO 检测网络的 FPGA 加速计算模型的研究 [J]. 小型微型计算机系统,2022,43(8):1681-1686.
[10] 胡晶晶 . 基于 FPGA 的 Faster-RCNN 改进算法实现目标检测 [J]. 现代计算机,2021,27(30):82-87.
作者简介:吴昱昊(1996—),女,汉族,浙江上虞人,助教,硕士研究生,研究方向:深度学习、社会网络分析。