Table 1.
Comparison of various real-time pose estimation methods on the COCO-POSE 2017 data set.
| Method | AP | AP50 | AP75 | APM | APL | AR |
|---|---|---|---|---|---|---|
| Hourglass [26] | 63.0 | 85.7 | 68.9 | 58.0 | 70.4 | - |
| HGG [27] | 67.6 | 85.1 | 73.7 | 62.7 | 74.6 | - |
| OpenPOSE [28] | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | 66.5 |
| HRNet | 64.5 | 86.2 | 71.7 | 59.2 | 73.5 | - |
| HigherHRNet [29] | 70.5 | 89.6 | 76.9 | 65.1 | 76.9 | - |
| YOLOv5m6-rlepose [30] | 67.6 | 89.2 | 74.5 | - | 73.8 | 74.0 |
| YOLOv8-PoseBoost [31] | - | 85.4 | 59.1 | 60.1 | 74.2 | - |
| YOLOv8l-POSE | 67.6 | 90.0 | 74.6 | 63.3 | 75.7 | 74.9 |
| SDPose-T [32] | 69.2 | 90.2 | 76.8 | 65.7 | 75.2 | - |
| YOLOv8x-POSE | 71.7 | 91.5 | 78.4 | 68.2 | 78.3 | 77.9 |
| Diffusion RegPose [33] | 72.5 | 89.9 | 79.5 | 66.8 | 80.5 | - |
| Ours | 69.1 | 90.9 | 75.5 | 65.2 | 76.6 | 76.7 |