A CNN-based 3D human pose estimation based on projection of depth and ridge data

作者：

Highlights：

• We propose a CNN-based human pose estimation using depth and ridge data.

• We project the depth and ridge data on three orthogonal planes (XY, XZ, ZY).

• The projected depth and ridge data can eliminate the 3D information loss.

• Ridge data is introduced to avoid joint drift, which improves the accuracy of estimated poses.

• The proposed method achieved the state-of-the-art pose estimation accuracies.

摘要

•We propose a CNN-based human pose estimation using depth and ridge data.•We project the depth and ridge data on three orthogonal planes (XY, XZ, ZY).•The projected depth and ridge data can eliminate the 3D information loss.•Ridge data is introduced to avoid joint drift, which improves the accuracy of estimated poses.•The proposed method achieved the state-of-the-art pose estimation accuracies.

论文关键词：3D Human pose estimation,3D Point projection,Ridge data

论文评审过程：Received 26 September 2019, Revised 13 May 2020, Accepted 16 May 2020, Available online 20 May 2020, Version of Record 26 May 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107462