针对三维人体姿态估计实际应用场景需求,提出一种基于空洞卷积ResNet模块和稀疏自注意力(Sparse Attention,SA)的轻量化三维人体姿态估计模型DS-Net(Dilated Sparse Attention Network).首先,以单目、单阶段、多个三维人的回归网络(Monocular,One-stage,Regression of Multiple 3D People,ROMP)为基础姿态估计模型,并替换支路中基础ResNet模块的卷积为空洞卷积,在不降低精度的前提下减少模型参数量;其次,在支路中嵌入Sparse Attention,加强上下文理解能力以提高精度;最后,经过7个数据集训练和3DPW数据集测试,验证模型可行性.经实验验证,提出的DS-Net总参数量减少53.8%;在三维人体姿态估计任务中与ROMP相比,MPJPE和PA-MPJPE分别降低1.8%和2.9%,满足姿态估计实际应用场景需求.
A lightweight 3D human pose estimation algorithm
Aiming at the practical application requirements of 3D human pose estimation,a lightweight 3D human pose estimation model dilated sparse attention network based on dilated convolutional ResNet module and sparse attention is proposed.Firstly,the Monocular,One-stage,Regression of multiple 3D people network is used as the basic pose estimation model,and the convolution of the basic ResNet module in the branch is replaced by the dilated convolution.In order to reduce the number of model parameters without reducing the accuracy;Secondly,Sparse Attention is embedded in the branch to strengthen the context understanding ability to im-prove the accuracy.Finally,after 7 datasets training and 3DPW dataset testing,the feasibility of the model is verified.The experimental results show that the total number of DS-Net parameters is reduced by 53.8%.Compared with ROMP,MPJPE and PA-MPJPE reduce 1.8%and 2.9%respectively in 3D human pose estimation tasks,which can meet the requirements of practical application scenarios.
国家科技期刊平台
NSTL
万方数据
