학술논문
基于改进YOLOv7的滑雪摔倒检测 / Ski Fall Detection Based on Improved YOLOv7
Document Type
Academic Journal
Author
Source
现代信息科技 / Modern Informationn Technology. 8(1):84-88
Subject
Language
Chinese
ISSN
2096-4706
Abstract
针对目前滑雪场内滑雪人员摔倒检测存在的问题,提出一种基于YOLOv7的目标改进模型.对于检测模型部署在巡逻机器人上致使计算资源受限的问题,在主干网络中引入Ghost模型并在颈部引入GSConv降低模型参数;同时,引入基于并行可变形卷积的注意力机制模块(Parallel Deformable Attention Conv,PDAC)增强模型的精度.改进后的模型相较于原模型在参数上降低了21.6%,GFLOPs降低了27.7%,所需要的计算资源也大大降低.
A target improvement model based on YOLOv7 is proposed to address the current issues in detecting falls among skiers in ski resorts.For the problem of limited computing resources caused by deploying detection models on patrol robots,the Ghost model is introduced into the backbone network and GSConv is introduced in the neck to reduce model parameters;meanwhile,the Parallel Deformable Attention Conv(PDAC)module is introduced to enhance the accuracy of the model.The improved model has reduced parameters by 21.6%and GFLOPs by 27.7%compared to the original model,and the required computational resources have also been greatly reduced.
A target improvement model based on YOLOv7 is proposed to address the current issues in detecting falls among skiers in ski resorts.For the problem of limited computing resources caused by deploying detection models on patrol robots,the Ghost model is introduced into the backbone network and GSConv is introduced in the neck to reduce model parameters;meanwhile,the Parallel Deformable Attention Conv(PDAC)module is introduced to enhance the accuracy of the model.The improved model has reduced parameters by 21.6%and GFLOPs by 27.7%compared to the original model,and the required computational resources have also been greatly reduced.