摘要
为进一步探究人体姿势估计方面的准确度提升路径,本文以深度卷积神经网络为核心,首先,设计其整体框架;其次,综合考虑效率和准确度要求,从全局和细节两个层面着手设计模型;最后,初步测试该模型。测试结果表明,相较于当前主流的几种模型,该模型在准确度上具有一定优势。
To further explore the accuracy improvement path in human pose estimation,this paper focuses on deep convolutional neural networks.Firstly,the overall framework is designed.Secondly,considering the requirements of efficiency and accuracy comprehensively,the model is designed from both the global and detailed levels.Finally,conduct preliminary testing of the model.The test results indicate that compared to several mainstream models,the modified model has certain advantages in accuracy.
作者
吴华芹
柳静
马艺嘉
WU Huaqin;LIU Jing;MA Yijia(Henan Technical Institute,Zhengzhou,Henan 450042,China;Xi'an University of Architecture and Technology,Xi'an,Shaanxi 710055,China)
出处
《自动化应用》
2023年第9期36-38,44,共4页
Automation Application
基金
河南应用技术职业学院“首席技师”资助项目(2020-SXJS-XX04)
河南应用技术职业学院第一批课程思政示范课程项目《Java程序设计基础》
2022年河南省职业教育和继续教育课程思政示范课程《Java程序设计基础》
河南应用技术职业学院2020年“教学工程”项目Java程序设计教学创新团队。
关键词
深度卷积
神经网络
人体姿势估计
depth convolution
neural network
human posture estimation
