计算机应用 ›› 2021, Vol. 41 ›› Issue (2): 517-522.DOI: 10.11772/j.issn.1001-9081.2020050622

所属专题: 多媒体计算与计算机仿真

• 多媒体计算与计算机仿真 • 上一篇    下一篇

基于单点多盒检测器的全局-局部层级的域适应目标检测

蒋宁1,2, 方景龙1, 杨庆3   

  1. 1. 杭州电子科技大学 计算机学院, 杭州 310018;
    2. 宁波城市职业技术学院 信息与智能工程学院, 浙江 宁波 315110;
    3. 南京工程学院 计算机工程学院, 南京 211167
  • 收稿日期:2020-05-23 修回日期:2020-08-03 出版日期:2021-02-10 发布日期:2020-08-26
  • 通讯作者: 方景龙
  • 作者简介:蒋宁(1974-),男,浙江宁波人,高级系统分析师,博士研究生,主要研究方向:人工智能、虚拟现实;方景龙(1964-),男,江西景德镇人,教授,博士,主要研究方向:人工智能、机器学习;杨庆(1976-),女,江苏南京人,副教授,博士,主要研究方向:虚拟仿真、图形图像。
  • 基金资助:
    浙江省教育厅科研项目(Y201533716);江苏省科技计划项目(BY2016008-02)。

Global-local domain adaptive object detection based on single shot multibox detector

JIANG Ning1,2, FANG Jinglong1, YANG Qing3   

  1. 1. School of Computer Science and Technology, Hangzhou Dianzi University, Hangzhou Zhejiang 310018, China;
    2. School of Information and Intelligent Engineering, Ningbo City College of Vocational Technology, Ningbo Zhejiang 315110, China;
    3. School of Computer Engineering, Nanjing Institute of Technology, Nanjing Jiangsu 211167, China
  • Received:2020-05-23 Revised:2020-08-03 Online:2021-02-10 Published:2020-08-26
  • Supported by:
    This work is partially supported by the Scientific Research Project of Education Department of Zhejiang Province (Y201533716), the Science and Technology Program of Jiangsu Province (BY2016008-02).

摘要: 在目标检测领域里通常希望在拥有大量标记的场景中训练好的模型能够应用在无标记的其他场景中,但是不同的域分布往往是不同的,这样往往导致域迁移时模型性能的急剧下降。为了提高域迁移时模型的目标检测性能,通过两个层级来解决域迁移问题,包括全局层级迁移和局部层级迁移。这两种层级迁移分别对应不同的特征对齐方式,即全局层级采用选择性对齐方式,局部层级采用完全对齐方式。所提域迁移框架基于单点多盒检测器(SSD)模型,在全局和局部层级分别配置相应的域适配器以减少域间差异,通过对抗网络算法实现具体训练,再通过一致性正则化来进一步提高模型的域迁移性能。通过大量实验验证了提出的域迁移模型的有效性,结果表明同目前常见的域适应-快速区域卷积(DA-FRCNN)模型、对抗识别域适应(ADDA)模型以及动态对抗适应网络(DAAN)模型等三种域迁移模型相比,该模型在不同数据集上的均值平均精度(mAP)可以提高5%~10%。

关键词: 目标检测, 域适配器, 域迁移, 对抗训练, 特征对齐, 卷积层, 损失函数, 单点多盒检测器

Abstract: In the field of object detection, it is hoped that the model trained in the domain with a lot of labels can be applied to other domains without labels, but different domain distributions are always different to each other, such difference will result in a sharp decline of model performance in domain transfer. To improve the model performance of object detection in domain transfer, the domain transfer was addressed on two levels, including the global-level transfer and the local-level transfer, which were corresponding to different feature alignment methods, that is, the global-level adopted selective alignment and the local-level adopted full alignment. The proposed domain transfer framework was constructed based on Single Shot MultiBox Detector (SSD) model and was disposed of two domain adaptors corresponding to global and local level respectively for the purpose of alleviating the domain difference. The specific training was implemented by the adversarial network algorithm, and the consistency regularization was used to further improve the domain transfer performance of the model. The effectiveness of the proposed domain transfer model was verified by many experiments. Experimental results show that on various datasets, the proposed model outperforms the existing common domain transfer models such as Domain Adaptation-Faster Region-based Convolutional Neural Network(DA-FRCNN), Adversarial Discriminative Domain Adaptation (ADDA), Dynamic Adversarial Adaptation Network (DAAN) by 5%-10% in term of mean Average Precision (mAP).

Key words: object detection, domain adaptor, domain transfer, adversarial training, feature alignment, convolution layer, loss function, Single Shot MultiBox Detector (SSD)

中图分类号: