[1]徐伟峰,雷耀,王洪涛,等.面向边缘设备的目标检测模型研究[J].智能系统学报,2025,20(4):871-881.[doi:10.11992/tis.202406015]
XU Weifeng,LEI Yao,WANG Hongtao,et al.Research on object detection models for edge devices[J].CAAI Transactions on Intelligent Systems,2025,20(4):871-881.[doi:10.11992/tis.202406015]
点击复制
XU Weifeng,LEI Yao,WANG Hongtao,et al.Research on object detection models for edge devices[J].CAAI Transactions on Intelligent Systems,2025,20(4):871-881.[doi:10.11992/tis.202406015]
面向边缘设备的目标检测模型研究
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
20
期数:
2025年第4期
页码:
871-881
栏目:
学术论文—机器学习
出版日期:
2025-08-05
- Title:
- Research on object detection models for edge devices
- Keywords:
- object detection; YOLO; edge devices; inference accuracy; inference speed; data read/write volume; computational load; model deployment
- 分类号:
- TP391.4
- 文献标志码:
- 2024-12-12
- 摘要:
- 现有目标检测模型在边缘设备上部署时，其检测性能和推理速度的平衡有较大提升空间。针对此问题，本文基于YOLO (you can only look once) v8提出一种可部署到多类边缘设备上的目标检测模型。在模型的骨干网络部分，设计了EC2f (extended coarse-to-fine) 结构，在降低参数量和计算复杂度的同时降低数据读写量；在颈部网络部分，将颈部网络替换为YOLOv6-3.0版本的颈部网络，加速了模型推理，并将推理精度维持在较好水平；预测头网络部分设计了多尺度卷积检测头，进一步降低了模型的计算复杂度和参数度。设计了两个版本 (n/s尺度)以适应不同的边缘设备。在X光数据集的实验表明，模型在推理精度上比同尺度的基准模型分别提升0.5/1.7百分点，推理速度上分别提升11.6%/11.2%。在其他数据集上的泛化性能测试表明，模型的推理速度提升了10%以上，精度降低控制在1.3%以内。实验证明，模型在推理精度和速度之间实现了良好的平衡。
- Abstract:
- Existing object detection models can be improved in terms of balancing detection performance and inference speed on edge devices. Hence, a YOLO (you can only look once) v8-based model optimized for various edge devices is proposed. In the Backbone, an EC2f (extended coarse-to-fine) structure is designed to reduce parameters, computation, and data read/write volume. In the Neck, the YOLOv6-3.0 version is used to accelerate inference while maintaining accuracy. In the Head, a multiscale convolutional detection head, which further reduces computational load and complexity, is featured. Two versions (n/s scales) are designed to suit different edge devices. Experiments on an X-ray dataset demonstrate that the proposed model improves inference accuracy by 0.5%/1.7% and speed by 11.6%/11.2% compared with baseline models of the same scale. Generalization tests on other datasets present an increase in inference speed of over 10% and an accuracy reduction controlled within 1.3%. Overall, the model achieves a satisfactory balance between inference accuracy and speed.
- 参考文献/References:
-
[1] REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(6): 1137-1149.
[2] HE Kaiming, GKIOXARI G, DOLLáR P, et al. Mask R-CNN[C]//2017 IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2980-2988.
[3] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.
[4] LIU Wei, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[M]//Lecture Notes in Computer Science. Cham: Springer International Publishing, 2016: 21-37.
[5] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//2017 IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2999-3007.
[6] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 779-788.
[7] REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-08)[2024-06-11]. https://arxiv.org/abs/1804.02767v1.
[8] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23)[2024-06-11]. https://arxiv.org/abs/2004.10934v1.
[9] DUAN Kaiwen, BAI Song, XIE Lingxi, et al. CenterNet: keypoint triplets for object detection[C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 6568-6577.
[10] TIAN Zhi, CHU Xiangxiang, WANG Xiaoming, et al. Fully convolutional one-stage 3d object detection on lidar range images[J]. Advances in neural information processing systems, 2022, 35: 34899-34911.
[11] GE Zheng, LIU Songtao, WANG Feng, et al. Yolox: exceeding yolo series in 2021[EB/OL]. (2021-07-18)[2024-06-11]. https://arxiv.org/abs/2107.08430.
[12] JOCHER G, CHAURASIA A, QIU J. Ultralytics YOLO (Version 8.0. 0). (2023-01-15)[2024-06-11]. http://github.com/ultralytics/ultralytics.
[13] HOWARD A G, ZHU Menglong, CHEN Bo, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[EB/OL]. (2017-04-17)[2024-06-11]. https://arxiv.org/abs/1704.04861v1.
[14] ZHANG Xiangyu, ZHOU Xinyu, LIN Mengxiao, et al. ShuffleNet: an extremely efficient convolutional neural network for mobile devices[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 6848-6856.
[15] HAN Kai, WANG Yunhe, TIAN Qi, et al. GhostNet: more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 1577-1586.
[16] JOCHER G. YOLOv5 by Ultralytics (Version 7.0). [2024-06-11]. https://doi.org/10.5281/zenodo.3908559.
[17] LI Chuyi, LI Lulu, GENG Yifei, et al. YOLOv6 v3.0: a full-scale reloading[EB/OL]. (2023-01-13)[2024-06-11]. https://arxiv.org/abs/2301.05586v1.
[18] WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver: IEEE, 2023: 7464-7475.
[19] 李源鑫, 郭忠峰, 杨钧麟. 基于轻量化YOLOv5s的集装箱锁孔识别算法[J]. 计算机科学, 2024, 51(S1): 524-529.
LI Yuanxin, GUO Zhongfeng, YANG Junlin. Container keyhole identification algorithm based on lightweight YOLOv5s[J]. Computer science, 2024, 51(S1): 524-529.
[20] 曲英伟, 刘锐. 基于YOLOv5-MobileNetV3算法的目标检测[J]. 计算机系统应用, 2024, 33(7): 213-221.
QU Yingwei, LIU Rui. Object detection based on YOLOv5-MobileNetV3 algorithm[J]. Computer systems and applications, 2024, 33(7): 213-221.
[21] 何宇豪, 易明发, 周先存, 等. 基于改进的Yolov5的无人机图像小目标检测[J]. 智能系统学报, 2024, 19(3): 635-645.
HE Yuhao, YI Mingfa, ZHOU Xiancun, et al. UAV image small-target detection based on improved Yolov5[J]. CAAI transactions on intelligent systems, 2024, 19(3): 635-645.
[22] TAN Mingxing, PANG Ruoming, LE Q V. EfficientDet: scalable and efficient object detection[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 10778-10787.
[23] 胡丹丹, 张忠婷. 基于改进YOLOv5s的面向自动驾驶场景的道路目标检测算法[J]. 智能系统学报, 2024, 19(3): 653-660.
HU Dandan, ZHANG Zhongting. Road target detection algorithm for autonomous driving scenarios based on improved YOLOv5s[J]. CAAI transactions on intelligent systems, 2024, 19(3): 653-660.
[24] GUPTA C, GILL N S, GULIA P, et al. A novel finetuned YOLOv6 transfer learning model for real-time object detection[J]. Journal of real-time image processing, 2023, 20(3): 42.
[25] CHEN Junyang, LIU Hui, ZHANG Yating, et al. A multiscale lightweight and efficient model based on YOLOv7: applied to citrus orchard[J]. Plants, 2022, 11(23): 3260.
[26] WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[M]//Lecture Notes in Computer Science. Cham: Springer International Publishing, 2018: 3-19.
[27] 邹珺淏, 任酉贵, 冷芳玲, 等. LW-YOLOv7SAR: 轻量SAR图像目标检测方法[J/OL]. 小型微型计算机系统, 1-9. [2024-06-15]. https://www.cnki.com.cn/Article/CJFDTotal-XXWX20231103009.htm.
ZOU Junhao, REN Yougui, LENG Fangling, et al. LW-YOLOv7SAR: Lightweight SAR image object detection method [J/OL]. Journal of Small Computer Systems, 1-9. [2024-06-15]. https://www.cnki.com.cn/Article/CJFDTotal-XXWX20231103009.htm.
[28] 高德勇, 陈泰达, 缪兰. 改进YOLOv8n的道路目标检测算法[J]. 计算机工程与应用, 2024, 60(16): 186-197.
GAO Deyong, CHEN Taida, MIAO Lan. Improved road object detection algorithm for YOLOv8n[J]. Computer engineering and applications, 2024, 60(16): 186-197.
[29] LIU Shu, QI Lu, QIN Haifang, et al. Path aggregation network for instance segmentation[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 8759-8768.
[30] YANG Guoyu, LEI Jie, ZHU Zhikuan, et al. AFPN: asymptotic feature pyramid network for object detection[C]//2023 IEEE International Conference on Systems, Man, and Cybernetics. Honolulu: IEEE, 2023: 2184-2189.
[31] CHEN Jierun, KAO S H, HE Hao, et al. Run, don’t walk: chasing higher FLOPS for faster neural networks[C]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver: IEEE, 2023: 12021-12031.
[32] DING Xiaohan, ZHANG Xiangyu, MA Ningning, et al. RepVGG: making VGG-style ConvNets great again[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville: IEEE, 2021: 13728-13737.
- 相似文献/References:
-
[1]胡光龙,秦世引.动态成像条件下基于SURF和Mean shift的运动目标高精度检测[J].智能系统学报,2012,7(1):61.
HU Guanglong,QIN Shiyin.High precision detection of a mobile object under dynamic imaging based on SURF and Mean shift[J].CAAI Transactions on Intelligent Systems,2012,7():61.
[2]韩峥,刘华平,黄文炳,等.基于Kinect的机械臂目标抓取[J].智能系统学报,2013,8(2):149.[doi:10.3969/j.issn.1673-4785.201212038]
HAN Zheng,LIU Huaping,HUANG Wenbing,et al.Kinect-based object grasping by manipulator[J].CAAI Transactions on Intelligent Systems,2013,8():149.[doi:10.3969/j.issn.1673-4785.201212038]
[3]韩延彬,郭晓鹏,魏延文,等.RGB和HSI颜色空间的一种改进的阴影消除算法[J].智能系统学报,2015,10(5):769.[doi:10.11992/tis.201410010]
HAN Yanbin,GUO Xiaopeng,WEI Yanwen,et al.An improved shadow removal algorithm based on RGB and HSI color spaces[J].CAAI Transactions on Intelligent Systems,2015,10():769.[doi:10.11992/tis.201410010]
[4]曾宪华,易荣辉,何姗姗.流形排序的交互式图像分割[J].智能系统学报,2016,11(1):117.[doi:10.11992/tis.201505037]
ZENG Xianhua,YI Ronghui,HE Shanshan.Interactive image segmentation based on manifold ranking[J].CAAI Transactions on Intelligent Systems,2016,11():117.[doi:10.11992/tis.201505037]
[5]葛园园,许有疆,赵帅,等.自动驾驶场景下小且密集的交通标志检测[J].智能系统学报,2018,13(3):366.[doi:10.11992/tis.201706040]
GE Yuanyuan,XU Youjiang,ZHAO Shuai,et al.Detection of small and dense traffic signs in self-driving scenarios[J].CAAI Transactions on Intelligent Systems,2018,13():366.[doi:10.11992/tis.201706040]
[6]莫宏伟,汪海波.基于Faster R-CNN的人体行为检测研究[J].智能系统学报,2018,13(6):967.[doi:10.11992/tis.201801025]
MO Hongwei,WANG Haibo.Research on human behavior detection based on Faster R-CNN[J].CAAI Transactions on Intelligent Systems,2018,13():967.[doi:10.11992/tis.201801025]
[7]宁欣,李卫军,田伟娟,等.一种自适应模板更新的判别式KCF跟踪方法[J].智能系统学报,2019,14(1):121.[doi:10.11992/tis.201806038]
NING Xin,LI Weijun,TIAN Weijuan,et al.Adaptive template update of discriminant KCF for visual tracking[J].CAAI Transactions on Intelligent Systems,2019,14():121.[doi:10.11992/tis.201806038]
[8]伍鹏瑛,张建明,彭建,等.多层卷积特征的真实场景下行人检测研究[J].智能系统学报,2019,14(2):306.[doi:10.11992/tis.201710019]
WU Pengying,ZHANG Jianming,PENG Jian,et al.Research on pedestrian detection based on multi-layer convolution feature in real scene[J].CAAI Transactions on Intelligent Systems,2019,14():306.[doi:10.11992/tis.201710019]
[9]刘召,张黎明,耿美晓,等.基于改进的Faster R-CNN高压线缆目标检测方法[J].智能系统学报,2019,14(4):627.[doi:10.11992/tis.201905026]
LIU Zhao,ZHANG Liming,GENG Meixiao,et al.Object detection of high-voltage cable based on improved Faster R-CNN[J].CAAI Transactions on Intelligent Systems,2019,14():627.[doi:10.11992/tis.201905026]
[10]单义,杨金福,武随烁,等.基于跳跃连接金字塔模型的小目标检测[J].智能系统学报,2019,14(6):1144.[doi:10.11992/tis.201905041]
SHAN Yi,YANG Jinfu,WU Suishuo,et al.Skip feature pyramid network with a global receptive field for small object detection[J].CAAI Transactions on Intelligent Systems,2019,14():1144.[doi:10.11992/tis.201905041]
备注/Memo
收稿日期:2024-6-11。
基金项目:国家自然科学基金项目（61802124）；中央高校基本科研业务费专项（2023MS137）；中国高校产学研创新基金项目（2023DT6）.
作者简介:徐伟峰，讲师，博士，主要研究方向为图像识别技术、形式化验证方法和低空空管系统，承担科研项目10项。E-mail：weifengxu@163.com。;雷耀，硕士研究生，主要研究方向为深度学习和目标检测，发表学术论文1篇。E-mail：2260140046@qq.com。;王洪涛，副教授，博士，中国计算机学会会员，主要研究方向为人工智能安全、自然语言处理、隐私计算和知识计算。主持国家自然科学基金项目1项、中央高校基本科研业务费专项2项。发表学术论文28篇。E-mail：wanght@ncepu.edu.cn。
通讯作者:王洪涛. E-mail：wanght@ncepu.edu.cn
更新日期/Last Update:
1900-01-01