《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (8): 2325-2329.DOI: 10.11772/j.issn.1001-9081.2022121865

• 第十九届CCF中国信息系统及应用大会 • 上一篇    

基于孪生网络的小样本目标检测算法

姜钧舰1, 刘达维1, 刘逸凡1, 任酉贵1,2, 赵志滨1()   

  1. 1.东北大学 计算机科学与工程学院,沈阳 110169
    2.辽宁省自然资源事务服务中心,沈阳 110001
  • 收稿日期:2022-12-15 修回日期:2023-02-02 接受日期:2023-02-08 发布日期:2023-04-21 出版日期:2023-08-10
  • 通讯作者: 赵志滨
  • 作者简介:姜钧舰(1998—),男,辽宁丹东人,硕士研究生,CCF会员,主要研究方向:机器学习、计算机视觉
    刘达维(1998—),男,辽宁沈阳人,硕士研究生,CCF会员,主要研究方向:机器学习、计算机视觉
    刘逸凡(1998—),男,河北保定人,硕士研究生,CCF会员,主要研究方向:机器学习、计算机视觉
    任酉贵(1981—),男,辽宁沈阳人,博士研究生,主要研究方向:遥感图像处理、计算机视觉;
  • 基金资助:
    国家自然科学基金资助项目(U1811261)

Few-shot object detection algorithm based on Siamese network

Junjian JIANG1, Dawei LIU1, Yifan LIU1, Yougui REN1,2, Zhibin ZHAO1()   

  1. 1.School of Computer Science and Engineering,Northeastern University,Shenyang Liaoning 110169,China
    2.Service Center of Natural Resource Affairs of Liaoning Province,Shenyang Liaoning 110001,China
  • Received:2022-12-15 Revised:2023-02-02 Accepted:2023-02-08 Online:2023-04-21 Published:2023-08-10
  • Contact: Zhibin ZHAO
  • About author:JIANG Junjian, born in 1998, M. S. candidate. His research interests include machine learning, computer vision.
    LIU Dawei, born in 1998, M. S. candidate. His research interests include machine learning, computer vision.
    LIU Yifan, born in 1998, M. S. candidate. His research interests include machine learning, computer vision.
    REN Yougui, born in 1981, Ph. D. candidate. His research interests include remote sensing image processing, computer vision.
  • Supported by:
    National Natural Science Foundation of China(U1811261)

摘要:

基于深度学习的目标检测算法如YOLO(You Only Look Once)和Faster R-CNN(Faster Region-Convolutional Neural Network)需要大量训练数据以保证模型的精度,而在很多场景下获取数据以及标注数据的成本较高;并且由于缺少海量的训练数据,导致检测的范围受限。针对以上问题,提出了一种基于孪生网络的小样本目标检测算法(SiamDet),旨在使用少量标注图像训练具有一定泛化能力的目标检测模型。首先,提出了基于深度可分离卷积的孪生网络,并使用深度可分离卷积设计了特征提取网络ResNet-DW,从而解决了样本不充足带来的过拟合问题;其次,基于孪生网络,提出了目标检测算法SiamDet,并在ResNet-DW的基础上,引入区域建议网络(RPN)来定位感兴趣目标;然后,引入二值交叉熵损失进行训练,并使用对比训练策略,从而增加了类别之间的区分度。实验结果表明,SiamDet在小样本条件下具有良好的目标检测能力,且相较于次优的算法DeFRCN(Decoupled Faster R-CNN),SiamDet在MS-COCO数据集20-way 2-shot和PASCAL VOC数据集5-way 5-shot上的AP50分别增加了4.1%和2.6%。

关键词: 目标检测, 小样本学习, 孪生网络, 深度可分离卷积, 对比训练

Abstract:

Deep learning based algorithms such as YOLO (You Only Look Once) and Faster Region-Convolutional Neural Network (Faster R-CNN) require a huge amount of training data to ensure the precision of the model, and it is difficult to obtain data and the cost of labeling data is high in many scenarios. And due to the lack of massive training data, the detection range is limited. Aiming at the above problems, a few-shot object Detection algorithm based on Siamese Network was proposed, namely SiamDet, with the purpose of training an object detection model with certain generalization ability by using a few annotated images. Firstly, a Siamese network based on depthwise separable convolution was proposed, and a feature extraction network ResNet-DW was designed to solve the overfitting problem caused by insufficient samples. Secondly, an object detection algorithm SiamDet was proposed based on Siamese network, and based on ResNet-DW, Region Proposal Network (RPN) was introduced to locate the interested objects. Thirdly, binary cross entropy loss was introduced for training, and contrast training strategy was used to increase the distinction among categories. Experimental results show that SiamDet has good object detection ability for few-shot objects, and SiamDet improves AP50 by 4.1% on MS-COCO 20-way 2-shot and 2.6% on PASCAL VOC 5-way 5-shot compared with the suboptimal algorithm DeFRCN (Decoupled Faster R-CNN).

Key words: object detection, few-shot learning, Siamese network, depthwise separable convolution, contrast training

中图分类号: