3D Human Pose Estimation on KTH Multiview Football II (Sequence 1 of Player 2)
100Upper Arms PCPgeometry-driven annotation harvesting approach
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| geometry-driven annotation harvesting approachCamera setup=Three cameras, Training data used=None from this dataset2017.04 | 100 | 100 | 100 | 100 | 100 | |
| [34]Discretize state space=true, Provided with 3D bounding box=true, Multi-person estimation=false2021.01 | 100 | 100 | 100 | 100 | 100 | |
| Iterative Greedy MatchingTrack smoothing=true2021.01 | 100 | 100 | 99 | 99.7 | 99.7 | |
| Iterative Greedy MatchingTrack smoothing=false2021.01 | 99 | 99 | 98 | 93 | 97 | |
| geometry-driven annotation harvesting approachCamera setup=Two cameras, Training data used=None from this dataset2017.04 | 98 | 92 | 99 | 97 | 96.5 | |
| Belagiannis et al. [5]Camera setup=Three cameras2017.04 | 98 | 72 | 99 | 92 | 90.3 | |
| [4]2021.01 | 98 | 72 | 99 | 92 | 90 | |
| [13]2021.01 | 97 | 95 | 100 | 98 | 98 | |
| Belagiannis et al. [5]Camera setup=Two cameras2017.04 | 96 | 68 | 98 | 88 | 87.5 | |
| Kazemi et al.Camera setup=Three cameras2017.04 | 89 | 68 | 100 | 99 | 89 | |
| [23]Discretize state space=true, Provided with 3D bounding box=true, Multi-person estimation=false2021.01 | 89 | 68 | 100 | 99 | 89 | |
| Belagiannis et al. [4]Camera setup=Three cameras2017.04 | 68 | 56 | 78 | 70 | 68 | |
| [3]2021.01 | 68 | 56 | 78 | 70 | 68 | |
| Belagiannis et al. [4]Camera setup=Two cameras2017.04 | 64 | 50 | 75 | 66 | 63.8 | |
| Burenius et al.Camera setup=Three cameras2017.04 | 60 | 35 | 100 | 90 | 71.2 | |
| [7]Discretize state space=true, Provided with 3D bounding box=true, Multi-person estimation=false2021.01 | 60 | 35 | 100 | 90 | 71 | |
| Burenius et al.Camera setup=Two cameras2017.04 | 53 | 28 | 88 | 82 | 62.7 |