高分辨率特征保持的头部姿态软阶段回归算法  

Head pose estimation based high-resolution feature maintained soft-stage regression

作　　者：莫建文[1] 梁豪昌 袁华[1] 姜贵昀 陈明瑶 Mo Jianwen;Liang Haochang;Yuan Hua;Jiang Guiyun;Chen Mingyao(School of Information and Communication,Guilin University of Electronic Technology,Guilin 541004,China;Guilin Yuanwang Intelligent Communication Technology Co.,Guilin 541004,China)

机构地区：[1]桂林电子科技大学信息与通信学院,桂林541004 [2]桂林远望智能通信科技有限公司,桂林541004

出　　处：《电子测量技术》2024年第7期130-137,共8页Electronic Measurement Technology

基　　金：国家自然科学基金(62177012,62001133);广西科技重大专项(桂科AA20302001)资助。

摘　　要：针对在头部姿态估计推理过程中由于上下采样操作而导致的姿态特征损失问题,提出了一种高分辨率特征保持的头部姿态软阶段回归算法。该算法首先利用编码器HR-Net对原始人脸图像进行高分辨率特征保持的多尺度特征编码,并在其卷积块中加入TA维度交互模块以捕获更多空间与通道之间的交互信息;然后使用解码器SSR-Net算法对HR-Net输出的不同尺度特征图进行关键参数解码和头部姿态软阶段回归,并引入了高效通道注意力ECA以加强特征通道间的信息交互,减少冗余特征。实验结果表明,所提算法在公开数据集AFLW2000和BIWI上均有优秀表现,其MAE分别降低至4.19和3.00。Aiming at the problem of pose feature loss due to up and down sampling in the inference process of head pose estimation,a high-resolution feature maintained soft-stage regression algorithm for head pose estimation is proposed.The algorithm first utilizes the encoder HR-Net to encode multiscale features for high-resolution feature maintaining in raw face images,and TA dimension interaction module joined in its convolutional block to capture more spatial-channel interaction information.The decoder SSR-Net algorithm was then applied to decode the key parameters and soft-stage regression of head pose on the different scale features output from HR-Net,and the Efficient Channel Attention ECA is employed to enhance the information interaction between feature channels and reduce redundant features.The experimental results show that the proposed algorithm has excellent performance on both the public datasets AFLW2000 and BIWI,and its MAE is reduced to 4.19 and 3.00,respectively.

关 键 词：头部姿态估计 高分辨率特征 软阶段回归 信息交互 TA维度交互 ECA注意力 

分 类 号：TN911.73[电子电信—通信与信息系统]