基于YOLOv8的轻量化道路裂缝检测模型
A Lightweight Road Crack Detection Model Based on YOLOv8
任晶晶 1徐志远1
作者信息
- 1. 太原学院智能与自动化系,山西 太原 030032
- 折叠
摘要
道路裂缝是主要的路面病害之一,及时、有效地检测道路裂缝对路面养护和管理非常重要.为减少模型参数量,提高推理速度,提出一种基于YOLOv8 改进的轻量化道路裂缝检测模型.该模型通过在YOLOv8 网络中的Neck层嵌入轻量化模块FasterNet,减少冗余计算和内存访问的同时,可以更有效的提取空间特征.在自制道路裂缝数据集上进行实验验证算法的检测效果,实验结果表明,改进后的模型可大幅减少模型参数和计算量,而且Recall和MAP均有一定程度的提升,在保证路面裂缝检测精度的同时也便于在嵌入式设备中部署.
Abstract
Road cracks are one of the main pavement diseases,and timely and effective detection of road cracks is crucial for pavement maintenance and management.To reduce the model parameter size,improve infer-ence speed,a lightweight road crack detection model based on improved YOLOv8 is proposed.This model embeds a lightweight module called FasterNet into the Neck layer of the YOLOv8 network,that reducing redundant compu-tations and memory access while effectively extracting spatial features.Experimental verification of the algorithm's detection performance is conducted on a self-made road crack dataset.The results show that the improved model significantly reduces model parameters and computational complexity,while also improving Recall and MAP to a certain extent.It ensures the accuracy of road crack detection and facilitates deployment on embedded devices.
关键词
道路裂缝/目标检测/YOLOv8/轻量化Key words
road crack/target detection/YOLOv8/lightweight引用本文复制引用
基金项目
山西省高等学校教学改革创新项目(J20221194)
大学生创新创业训练计划项目(TYX2022034)
出版年
2024
NETL
NSTL
万方数据