[1]刘雅楠,李维乾.融合 ECA 机制的轻量化 YOLOv4 检测模型[J].计算机技术与发展,2023,33(07):146-153.[doi:10. 3969 / j. issn. 1673-629X. 2023. 07. 022]
　LIU Ya-nan,LI Wei-qian.Lightweight YOLOv4 Detection Model Incorporating ECA Mechanism[J].,2023,33(07):146-153.[doi:10. 3969 / j. issn. 1673-629X. 2023. 07. 022]

点击复制

融合 ECA 机制的轻量化 YOLOv4 检测模型()

分享到：

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

卷:

33

期数:

2023年07期

页码:

146-153

栏目:

人工智能

出版日期:

2023-07-10

文章信息/Info

Title:

Lightweight YOLOv4 Detection Model Incorporating ECA Mechanism

文章编号:

1673-629X(2023)07-0146-08

作者:

刘雅楠; 李维乾

西安工程大学 计算机科学学院,陕西 西安 710048

Author(s):

LIU Ya-nan; LI Wei-qian

School of Computer Science,Xi’an Polytechnic University,Xi’ an 710048,China

关键词:

目标检测; YOLOv4; GhostNet; 轻量化神经网络; 注意力机制

Keywords:

object detection; YOLOv4; GhostNet; lightweight neural network; attention mechanism

分类号:

TP391. 4

DOI:

10. 3969 / j. issn. 1673-629X. 2023. 07. 022

摘要:

　　近年来,卷积神经网络已在人脸识别、无人驾驶等领域取得重大突破。 随着智能移动设备的普及,高精度的大型网络

往往伴随着参数量多、计算量大等问题,无法部署在这些资源有限的移动设备平台上。 GhostNet 通过简单的线性操作生

成更多特征映射,可大幅减少计算成本。 为此,提出了一种改进轻量化 YOLOv4 的 GhostNet-YOLOv4 网络模型,该模型将 YOLOv4 的主干网络替换为 GhostNet 残差结构,借助即插即用的 Ghost 模块升级卷积神经网络,并使用 Mosaic 数据增强

技术对数据集进行预处理,融合 ECA 机制,加入 Focal Loss 焦点损失函数,在保证一定精度的前提下大幅减少了模型的参数量和计算量。 相对于改进前的 GhostNet-YOLOv4 模型,改进后的 GhostNet-YOLOv4 在 PASCAL VOC 2007 数据集上的mAP ( mean Average Precision) 提高 1. 65 百分点,达到 85. 09% ,且模型参数量只有 11. 429 M,相对于原 YOLOv4 模型减少了约 80% ,具有更好的综合性能。

Abstract:

In recent years, convolutional neural networks have made breakthroughs in applications such as face recognition?

and　autonomous driving. With the popularity of intelligent mobile devices,high - precision large - scale networks are often accompanied byproblems such as a large number of parameters and a large amount of computing, and cannot?

be deployed on these resource - limitedmobile device platforms. GhostNet generates more feature maps with simple linear operations, drastically reducing the computationalcost. Therefore,we propose an improved lightweight YOLOv4 network model. The backbone network of YOLOv4 is replaced with theGhostNet residual structure,which can upgrade the convolutional neural networks, and the Mosaic data enhancement is used to pre -process the dataset. ECA mechanism is integrated and Focal Loss function is added. On the premise of ensuring certain accuracy,thenumber of parameters and calculation amount of the model are greatly reduced. Compared with the GhostNet-YOLOv4 before the improvement,the mAP ( mean Average Precision) of the improved GhostNet-YOLOv4 on the PASCAL VOC 2007 dataset is increased by1. 65% to 85. 09% , and the parameters of the model are only 11. 429 M, which reduces by about 80% compared with the originalYOLOv4 network,which indicates that the improved model has better overall performance.

常用功能