A lightweight deep learning network RepYOLO for embedded devices

doi:10.14088/j.cnki.issn0439-8114.2024.08.026

Abstract

Abstract: A lightweight deep learning network model RepYOLO algorithm was proposed and transplanted to embedded device MCU/MPU. The network model RepYOLO took YOLOv4 as the base network model. By modifying YOLOv4’s backbone network CSPDarkNet to the RepBlock structure, introducing the CBAM attention mechanism in the Neck layer, and replacing the anchor-based detection head with an anchor-free detection head in the head layer along with integrating the ATSS algorithm, the computational load was reduced, and both inference speed and detection accuracy were improved. The experimental results showed that compared with the original YOLOv4 model, the network model RepYOLO showed more significant advantages in wheat spike detection, and its precision rate, recall rate, F1 value and average precision value were increased by 4.7, 3.6, 1.5 and 1.7 percentage points, respectively. In addition, RepYOLO reduced inference time on embedded devices MCU/MPU by 37.03% and 41.44%, respectively.

Key words: object detection, deep learning, embedded device, lightweight network, RepYOLO, wheat spike

CLC Number:

ZHOU Gan-wei, CHEN Jia-yue, WU Jia-wei, ZHAO Ya-qi, ZHAO Yi-kai, ZHANG Xiao-ying. A lightweight deep learning network RepYOLO for embedded devices[J]. HUBEI AGRICULTURAL SCIENCES, 2024, 63(8): 153-157.

References

[1] 吕志远,张付杰,魏晓明,等.采用组合增强的YOLOX-ViT协同识别温室内番茄花果[J].农业工程学报,2023,39(4):124-134.
[2] 张振国,邢振宇,赵敏义,等.改进YOLOv3的复杂环境下红花丝检测方法[J].农业工程学报,2023,39(3):162-170.
[3] 张日红,区建爽,李小敏,等.基于改进YOLOv4的轻量化菠萝苗心检测算法[J].农业工程学报,2023,39(4):135-143.
[4] 欧阳雪,徐彦彦,毛养素,等.云计算与区块链平台的遥感影像安全检索方案[J].电子与信息学报,2023,45(3):856-864.
[5] 李丹,邓飞,赵良玉,等.基于深度学习的无人机自主降落标识检测方法[J].航空兵器,2023,30(5):115-120.
[6] DAVID E,MADEC S,SADEGHI-TEHRAN P, et al.Global wheat head detection (GWHD) dataset: A large and diverse dataset of high-resolution RGB-labelled images to develop and benchmark wheat head detection methods[J]. Plant phenomics, 2020, 2020:3521852.
[7] 章倩丽,李秋生,胡俊勇,等.基于PP-YOLO改进算法的脐橙果实实时检测[J].北京联合大学学报,2022,36(4):58-66.
[8] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: Optimal speed and accuracy of object detection[J].[2023-06-10].http://arxiv.org/PDF/2004.10934.
[9] ZHANG Z, LU X, CAO G, et al.ViT-YOLO: Transformer-based YOLO for object detection[A]. 2021 IEEE/CVF international conference on computer vision workshops (ICCVW)[C].IEEE, 2021.2799-2808.
[10] WANG X, SONG J.ICIoU: Improved loss based on complete intersection over union for bounding box regression[J]. IEEE Access, 2021.105686-105695.
[11] 江明哲. 基于深度学习的复杂环境下番茄目标检测与定位技术研究[D].重庆:重庆邮电大学,2022.
[12] WANG K, LIEW J H, ZOU Y, et al.Panet: Few-shot image semantic segmentation with prototype alignment[A].2021 IEEE/CVF international conference on computer vision workshops (ICCVW)[C].IEEE, 2019.9197-9206.
[13] WOO S, PARK J,LEE J Y, et al.Cbam: Convolutional block attention module[A].2021 IEEE/CVF international conference on computer vision workshops (ICCVW)[C].IEEE, 2018. 3-19.
[14] GE Z, LIU S, WANG F, et al.Yolox: Exceeding yolo series in 2021[J]. arXiv, 2021, 2107:08430.
[15] ZHANG S, CHI C, YAO Y, et al.Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[A]. 2020 IEEE/CVF conference on computer vision and pattern recognition (CVPR)[C]. IEEE, 2020.9759-9768.
[16] 肖天元,艾廷华,余华飞,等.地图综合图卷积神经网络点群简化方法[J].测绘学报,2024,53(1): 158-172.
[17] 夏雪,柴秀娟,张凝,等.用于边缘计算设备的果树挂果量轻量化估测模型[J].智慧农业(中英文),2023,5(2): 1-12.
[18] ZHANG X, ZHOU X, LIN M, et al.Shufflenet: An extremely efficient convolutional neural network for mobile devices[C]. 2018 IEEE conference on computer vision and pattern recognition(CVPR)[C].IEEE, 2018.6848-6856.
[19] MA N, ZHANG X, ZHENG H T, et al.Shufflenet v2: Practical guidelines for efficient cnn architecture design[A]. 15th European conference on computer vision (ECCV)[C]. Springer, 2018.122-138.
[20] 赵迪, 叶盛波, 周斌.基于Grad-CAM的探地雷达公路地下目标检测算法[J].电子测量技术,2020,43(10): 113-118.