3D Human Pose Estimation on CMU Panoptic 17-subset (val)
13.7MPJPEVolumetric Triangulation
Evaluation Results
| Method | Links | |
|---|---|---|
| Volumetric TriangulationNumber of cameras=4, Aggregation method=softmax aggregation, 2D detector=Mask R-CNN, 2D detector backbone=ResNet-1522019.05 | 13.7 | |
| Volumetric TriangulationNumber of cameras=4, Aggregation method=sum aggregation, 2D detector=Mask R-CNN, 2D detector backbone=ResNet-1522019.05 | 13.7 | |
| Volumetric TriangulationNumber of cameras=4, Aggregation method=confidence aggregation, 2D detector=Mask R-CNN, 2D detector backbone=ResNet-1522019.05 | 14 | |
| Algebraic TriangulationNumber of cameras=4, Confidence usage=with confidence, 2D detector=Mask R-CNN, 2D detector backbone=ResNet-1522019.05 | 21.3 | |
| Algebraic TriangulationNumber of cameras=4, Confidence usage=without confidence, 2D detector=Mask R-CNN, 2D detector backbone=ResNet-1522019.05 | 33.4 | |
| RANSACNumber of cameras=4, 2D detector=Mask R-CNN, 2D detector backbone=ResNet-1522019.05 | 39.5 |