Semantic Graph Convolutional Networks for 3D Human Pose Regression

CVPR19

从2d坐标估计3d坐标, 这个参照17年ICCV A simple yet effective baseline for 3d human pose
estimation, 它证明了在2d坐标足够精确的情况下, 是可以估计出较为精确的3d坐标的。

本文结合了GCN, 加上Non-local layer。

Semantic Graph Convolutional Networks for 3D Human Pose Regression_第1张图片

Method 2D Detections # of Epochs # of Parameters MPJPE (P1) P-MPJPE (P2)
Martinez et al. [1] Ground truth 200 4.29M 44.40 mm 35.25 mm
SemGCN Ground truth 50 0.27M 42.14 mm 33.53 mm
SemGCN (w/ Non-local) Ground truth 30 0.43M 40.78 mm 31.46 mm
Martinez et al. [1] SH (fine-tuned) 200 4.29M 63.48 mm 48.15 mm
SemGCN (w/ Non-local) SH (fine-tuned) 100 0.43M 61.24 mm 47.71 mm

 我觉得效果其实挺一般的, 加了GCN等结构, 也只比ICCV17那片最简单的感知机模型好一点。

你可能感兴趣的:(Object,Detection)