YOLO系列模型是单阶段目标检测代表性算法,自2016年提出后历经五次重大迭代。本文系统分析YOLOv1至v5技术演进路径,从网络架构设计、损失函数优化、特征融合策略等维度提炼发展规律,结合当前技术趋势为其未来发展提出建议。
Abstract
The YOLO series model is a representative algorithm for single-stage object detection, and it has undergone five major iterations since it was proposed in 2016. This paper systematically analyzes the technical evolution path of YOLO v1 to v5, and summarizes the development laws from the dimensions of network architecture design, loss function optimization and feature fusion strategy. Suggestions for its future development are put forward in combination with the current technical trend.
关键词
目标检测 /
YOLO /
深度学习 /
网络架构 /
损失函数
Key words
object detection /
YOLO /
deep learning /
network architecture /
loss function
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 李睿鑫,张应迁,吴嘉懿,等. YOLOv5改进综述[J].电脑知识与技术,2024,20(27):19-22.
[2] 王鑫杰,王吉平. YOLO目标检测算法综述[J].广西物理,2024,45(2):50-53.
[3] 徐彦威,李军,董元方,等. YOLO系列目标检测算法综述[J].计算机科学与探索,2024,18(9):2221-2238.