Error Compensation Heatmap Decoding for Human Pose Estimation

IEEE Access

Jin Wu¹  Zhang Yao²  Pan Zhe²  Chen Yu²  Mo Yaoyang²  Zhang Min²  Qian Beibei²  Xiao Zhenzhong²  Song Zhan³  Yang Feiyu³ 

[1] College of Energy and Power Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, China;Orbbec Inc., Shenzhen, China;Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China;

关键词: Human pose estimation;    heatmap;    decoding;    error compensation;   

DOI  :  10.1109/ACCESS.2021.3105553

来源: DOAJ

【 摘 要 】

As a fundamental component of heatmap-based human pose estimation methods, heatmap decoding is to transform heatmaps into joint coordinates. We found that previous heatmap decoding methods generally ignored the effect of systematic errors introduced by the resolution increaseing operations in the network decoder. This work fills the gap by taking the systematic errors in heatmap decoding into consideration. We proposed a fast method to reduces the systematic and random errors in one shot by error compensation. The proposed method outperforms the previous best method on the COCO and the MPII datasets while being over 2 times faster. Extensive experiments with different networks, resolutions, metrics and datasets have proved the rationality of the proposed idea.

【 授权许可】

Unknown