BSMH: Cross-dataset object detection based on box-separated multiple-head

被引:1
作者
Lu, Feng [1 ]
Xu, You Chun [1 ]
Qi, Yao [1 ]
Xie, De Sheng [1 ]
Le Li, Yogn [1 ]
机构
[1] Army Mil Transportat Univ, Inst Mil Transportat, Tianjin, Peoples R China
关键词
computer vision; object detection;
D O I
10.1049/ipr2.13152
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Cross-dataset object detection methods can adapt to the needs of rapid category expansion in object detection tasks. However, these methods are prone to generate dataset-aware errors with false alarm objects. This study is aimed to address these issues. A box-separated multiple-head module and box-separated loss function based on the YOLOv8 network are devised to achieve cross-dataset object detection. Additionally, a sameclass-aware fusion module to avoid gradient conflicts due to cross-category conflicts is developed. A multiple-head fusion module is devised to reduce the number of false alarm objects caused by dataset-aware errors. A global class-aware sampler is also designed to adapt to the impact of the imbalanced number of categories and training samples across datasets. The effectiveness of the box-separated multiple-head module is verified using cross-datasets built using the COCO, WiderFace, WiderPerson, and OpenImages V4 datasets. Extensive experiments demonstrate the efficiency and precision of the proposed method. Cross-dataset object detection methods can adapt to the needs of rapid category expansion in object detection tasks. However, these methods is prone to generate dataset-aware errors with false alarm objects. A box-separated multiple-head module and box-separated loss function based on the YOLOv8 network are devised to achieve cross-dataset object detection. image
引用
收藏
页码:3013 / 3027
页数:15
相关论文
共 39 条
[1]  
[Anonymous], tion, DOI DOI 10.1109/CVPR.2016.596
[2]  
Bochkovskiy A, 2020, PREPRINT, DOI 10.48550/ARXIV.2004.10934
[3]  
Cao K., 2023, RELATIONAL MULTITASK
[4]  
Crawshaw M., 2020, arXiv
[5]   The PASCAL Visual Object Classes Challenge: A Retrospective [J].
Everingham, Mark ;
Eslami, S. M. Ali ;
Van Gool, Luc ;
Williams, Christopher K. I. ;
Winn, John ;
Zisserman, Andrew .
INTERNATIONAL JOURNAL OF COMPUTER VISION, 2015, 111 (01) :98-136
[6]  
Ge Z., 2021, ARXIV, DOI [10.48550/ARXIV.2107.08430, 10.48550/arXiv.2107.08430]
[7]  
Han C., 2022, YOLOPV2 BETTER FASTE
[8]  
He KM, 2020, IEEE T PATTERN ANAL, V42, P386, DOI [10.1109/ICCV.2017.322, 10.1109/TPAMI.2018.2844175]
[9]  
Hu J., 2018, WIDER FACE PERSON CH
[10]  
Jocher G., 2022, YOLOv5 release v6.1