| 摘要: |
| 基于深度半监督学习的目标检测技术利用少量带标注信息的样本和大量无标注信息的样本进行模型训练，可减少对标注样本的依赖，提高准确性和效率。首先介绍了基于深度半监督学习的目标检测理论，依据损失函数和模式设计方式的不同对其方法进行了分类，然后基于MS-COCO和Pascal VOC数据集对典型方法进行了性能对比，最后分析了其挑战和发展趋势，旨在为相关研究提供参考。 |
| 关键词: 目标检测 深度半监督学习 半监督学习 深度学习 |
| DOI：10.20079/j.issn.1001-893x.240605008 |
|
| 基金项目:2023年重庆市教委科学技术研究重点项目（KJZD-K202312903）；2023年度重庆市研究生科研创新项目（CYS23778）；2024年度重庆市研究生科研创新项目(CYS240832) |
|
| Survey of Object Detection Technology Based on Deep Semi-supervised Learning |
| HE Zhijie,XIAO Wei,LIU Nanqing,GAO Jiabo,KE Xueliang,QU Naizhu |
| (1.PLA Army Logistics Academy,Chongqing 401311,China;2.Unit 31680 of PLA,Chongzhou 611233,China;3.School of Information Science and Technology,Southwest Jiaotong University,Chengdu 610031,China) |
| Abstract: |
| The object detection technology based on deep semi-supervised learning uses a small number of samples with labeled information and a large number of samples without labeled information for model training,which reduces the dependence on labeled samples and improves the accuracy and efficiency.In this paper,the theory of object detection based on deep semi-supervised learning is introduced,and the methods are classified according to the different loss functions and pattern designs.Then,the performance of typical methods is compared based on MS-COCO and Pascal VOC data sets.Finally,their challenges and development trends are analyzed in order to provide a reference for related researches. |
| Key words: object detection deep semi-supervised learning semi-supervised learning deep learning |
