Cross View Fusion for 3D Human Pose Estimation

3 Sep 2019Haibo QiuChunyu WangJingdong WangNaiyan WangWenjun Zeng

We present an approach to recover absolute 3D human poses from multi-view images by incorporating multi-view geometric priors in our model. It consists of two separate steps: (1) estimating the 2D poses in multi-view images and (2) recovering the 3D poses from the multi-view 2D poses... (read more)

PDF Abstract

Evaluation results from the paper


 SOTA for 3D Human Pose Estimation on Human3.6M (using extra training data)

     Get a GitHub badge
Task Dataset Model Metric name Metric value Global rank Uses extra
training data
Compare
3D Human Pose Estimation Human3.6M Cross View Fusion Average MPJPE (mm) 31.17 # 2
3D Human Pose Estimation Human3.6M Cross View Fusion + MPII Average MPJPE (mm) 26.21 # 1
3D Human Pose Estimation Total Capture Fusion-RPSM Average MPJPE (mm) 29 # 1
3D Human Pose Estimation Total Capture Single-RPSM Average MPJPE (mm) 41 # 3