Table 2 Comparison of different models on crowdpose and sportspose datasets for pose estimation Task.
From: Deep learning-based approaches for human pose estimation in interdisciplinary physics applications
Model | CrowdPose Dataset | SportsPose Dataset | ||||||
|---|---|---|---|---|---|---|---|---|
Accuracy | Recall | F1 Score | AUC | Accuracy | Recall | F1 Score | AUC | |
OpenPose16 | 82.45 | 80.32 | 81.89 | 84.67 | 78.54 | 76.12 | 77.89 | 80.03 |
AlphaPose51 | 84.32 | 82.67 | 83.45 | 86.78 | 80.12 | 77.89 | 79.23 | 82.54 |
DeepCut52 | 81.23 | 79.34 | 80.67 | 83.45 | 76.45 | 74.12 | 75.89 | 78.34 |
HRNet18 | 86.45 | 84.78 | 85.67 | 88.12 | 82.34 | 79.89 | 81.12 | 84.45 |
SimpleBaseline53 | 83.12 | 81.56 | 82.67 | 85.23 | 79.45 | 76.78 | 78.34 | 81.45 |
DARKPose54 | 85.23 | 83.78 | 84.67 | 87.34 | 81.67 | 79.12 | 80.54 | 83.78 |
ViTPose55 | 87.34 | 85.23 | 86.27 | 88.76 | 82.34 | 80.56 | 81.45 | 84.67 |
TokenPose56 | 87.89 | 85.89 | 86.67 | 89.12 | 83.01 | 81.12 | 82.34 | 85.23 |
MixSTE57 | 88.45 | 86.78 | 87.23 | 89.67 | 83.45 | 81.78 | 82.89 | 85.89 |
Ours (HSTPN) | 89.12 | 87.45 | 88.34 | 90.45 | 84.23 | 82.12 | 83.67 | 86.34 |
