摘要
现有的基于深度学习的目标检测模型中,由于复杂天气使得现有方法的精度大幅降低.因此为了有效地消除不同天气场景带来的域差异问题,提出一种多域动态平均教师模型.首先引入多域平均教师模块,为多个不同天气场景下目标域数据生成伪标签;然后引入基于学生网络的风格迁移模块,解决多域任务中学生网络对不同目标域泛化能力差的问题,可有效地减小源域与不同目标域之间的差异,提升学生网络对不同目标域的泛化能力;最后提出基于教师网络的动态过滤伪标签模块,根据教师网络对不同目标域的学习效果动态地调整过滤伪标签的阈值,提升每个目标域伪标签质量.在FoggyCityscapes&RainCityscapes和Dusk-rain&Night-rain数据集上的实验结果表明,所提模型分别获得了40.3%和31.4%的精度,在雨天、雾天和夜晚等多种复杂天气场景下都优于对比方法.
The performance of existing deep learning-based object detection models significantly degrades due to the influence of complex weather.To effectively eliminate the problem of domain differences caused by dif-ferent weather scenes,propose a multi-domain dynamic mean teacher model.First,introduce a multi-domain mean teacher module to generate pseudo-labels for target domain data in multiple different weather scenes.Then,introduce a student network-based style transfer module to solve the problem of weak generalization ability of student network to different target domains in multi-domain tasks.It can effectively reduce the dif-ference between source domain and different target domains and improve the generalization ability of the stu-dent network to different target domains.Finally,a teacher network-based dynamic filtering pseudo-label module is used to dynamically adjust the threshold values of filtering pseudo labels according to the learning effect of the teacher network on different target domains and improve the quality of pseudo labels for each tar-get domain.Experiments on FoggyCityscapes&RainCityscapes and Dusk-rain&Night-rain datasets show that the proposed model achieves 40.3%and 31.4%accuracy respectively,and outperforms other comparison methods in complex weather scenes such as rain,fog and night.
作者
刘袁缘
王超凡
王文斌
张浩宇
罗忠文
Liu Yuanyuan;Wang Chaofan;Wang Wenbin;Zhang Haoyu;Luo Zhongwen(School of Computer Science,China University of Geosciences(Wuhan),Wuhan 430074;National Engineering Research Center for Geographic Information System,China University of Geosciences(Wuhan),Wuhan 430074)
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2024年第3期388-398,共11页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金面上项目(62076227)
武汉市科技局基础前沿项目(2020010601012166)。
关键词
领域自适应
目标检测
多目标域
风格迁移
伪标签
domain adaptation
object detection
multi-target domains
style transfer
pseudo-labeling
作者简介
刘袁缘(1984-),女,博士,副教授,博士生导师,CCF会员,主要研究方向为计算机视觉、模式识别;王超凡(1998-),男,硕士研究生,主要研究方向为目标检测;王文斌(1997-),男,硕士研究生,主要研究方向为人脸表情识别;张浩宇(1997-),男,硕士研究生,主要研究方向为多模态情感分析;通信作者:罗忠文(1963-),男,硕士,教授,博士生导师,CCF会员,主要研究方向为机器人与机器智能.luozw@cug.edu.cn。
