基于注意力特征金字塔的轻量级目标检测算法
2021年电子技术应用第10期
赵义飞,王 勇
北京工业大学 信息学部,北京100124
摘要: 基于深度学习的目标检测算法因其模型复杂度和对计算能力的要求,难以部署在移动设备等低算力平台上。为了降低模型的规模,提出一种轻量级目标检测算法。该算法在自顶向下的特征融合的基础之上,通过添加注意力机制构建特征金字塔网络,以达到更细粒度的特征表达能力。该模型以分辨率为320×320的图像作为输入,浮点运算量只有0.72 B,并在VOC数据集上取得了74.2%的mAP,达到了与传统单阶段目标检测算法相似的精度。实验数据表明,该算法在保持了检测精度的同时显著降低了模型运算量,更适合低算力条件下的目标检测。
中图分类号: TN98;TP391
文献标识码: A
DOI:10.16157/j.issn.0258-7998.211320
中文引用格式: 赵义飞,王勇. 基于注意力特征金字塔的轻量级目标检测算法[J].电子技术应用,2021,47(10):33-37.
英文引用格式: Zhao Yifei,Wang Yong. Lightweight object detection algorithm based on attention feature pyramid network[J]. Application of Electronic Technique,2021,47(10):33-37.
文献标识码: A
DOI:10.16157/j.issn.0258-7998.211320
中文引用格式: 赵义飞,王勇. 基于注意力特征金字塔的轻量级目标检测算法[J].电子技术应用,2021,47(10):33-37.
英文引用格式: Zhao Yifei,Wang Yong. Lightweight object detection algorithm based on attention feature pyramid network[J]. Application of Electronic Technique,2021,47(10):33-37.
Lightweight object detection algorithm based on attention feature pyramid network
Zhao Yifei,Wang Yong
Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China
Abstract: Object detection algorithms based on deep learning are difficult to deploy on low computing power platforms such as mobile devices due to their complexity and computational demands. In order to reduce the scale of the model, this paper proposed a lightweight object detection algorithm. Based on the top-down feature fusion, the algorithm built a feature pyramid network by adding an attention mechanism to achieve more fine-grained feature expression capabilities. The proposed model took an image with a resolution of 320×320 as input and had only 0.72 B FLOPs, achieved 74.2% mAP on the VOC dataset and the accuracy is similar to traditional one-stage object detection algorithms. Experimental data shows that the algorithm significantly reduces the computational complexity of the model, maintains the accuracy, and is more suitable for object detection with low computing power.
Key words : object detection;feature pyramid;attention mechanism;lightweight algorithm
0 引言
目标检测是计算机视觉的关键组成部分之一,旨在探索统一框架下人类视觉认知过程的模拟和行人检测、人脸识别、文本检测等特定应用场景下视觉任务的完成。2012年,Krizhevsky等[1]提出的AlexNet将卷积神经网络应用在了图像分类算法之中并取得了惊人的效果,从此基于深度学习的卷积神经网络算法开始取代传统的基于人工特征的算法,成为了计算机视觉领域的主流研究方向。
目前基于深度学习的目标检测算法可分为单阶段检测算法和两阶段检测算法两类。单阶段目标检测算法以SSD[2]和Yolo[3-5]系列算法为代表,是一种通过在卷积神经网络提取的特征图上设置锚点,并对每个锚点上预设的不同大小和长宽比例的边界框进行检测的方法。两阶段目标检测算法以RCNN[6-8]系列算法为代表,先在特征图上采用额外步骤生成候选区域,再对候选区域进行检测。与单阶段算法相比,两阶段算法一般拥有更高的检测精度,但由于增加了额外的运算量,检测速度也相对较低。
本文详细内容请下载:http://www.chinaaet.com/resource/share/2000003778。
作者信息:
赵义飞,王 勇
(北京工业大学 信息学部,北京100124)
此内容为AET网站原创,未经授权禁止转载。