一种改进的非极大值抑制算法在 YOLOv3 模型中的应用研究

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名

邮箱

手机号码

标题

留言内容

验证码

Application of an Improved Non Maximum Suppression Algorithm in YOLOv3 Model

1.Suzhou University of Science and Technology, Suzhou Jiangsu 215009; 2.Hangzhou University of Electronic Science and Technology, Hangzhou Zhejiang 310000

摘要: YOLOv3目标检测模型在进行候选区域确定时大多使用非极大值抑制算法,这类算法在多目标检测任务可能出现重复检测等问题,针对这一问题,本文在前人研究的soft-NMS算法基础上进行优化,提出I-NMS算法,并将该算法应用到YOLOv3 算法中进行建模,通过实验对比验证表明,I-NMS算法有助于解决YOLOv3算法中的重复检测问题。
- 目标检测 /
- YOLOv3算法 /
- 非极大值抑制
Abstract: Most of the YOLOv3 target detection models use non maximum suppression algorithm when determining the candidate region. This kind of algorithm may have the problem of repeated detection in multi-target detection task. Aiming at this problem, this paper optimizes the soft NMS algorithm studied by predecessors and proposes I-NMS algorithm. The algorithm is applied to YOLOv3 algorithm for modeling, and the experimental results show that the algorithm is effective I-NMS algorithm is helpful to solve the problem of duplicate detection in yolov3 algorithm.
- Object detection /
- YOLOv3 algorithm /
- Non maximum suppression

[1]	王元元.离散数学教程[M].北京:高等教育出版社,2015.
[2]	屈婉玲,耿素云,张立昂.离散数学[M].北京:高等教育出版社,2015.