Research Article
Application of Human Posture Recognition Based on the Convolutional Neural Network in Physical Training Guidance
Table 1
PCKh comparison results on the MPII dataset.
| Method | Head | Shoulder | Elbow | Wrist | Buttock | Knee | Ankle | PCKh |
| Iterative error feedback | 95.5 | 91.5 | 81.6 | 72.3 | 82.6 | 73.0 | 66.3 | 79.5 | Simple stacked hourglass network model | 98.1 | 96.0 | 91.1 | 86.9 | 90.1 | 87.2 | 83.4 | 90.1 | Network model based on improved stacked hourglass | 98.3 | 96.2 | 91.7 | 88.0 | 90.5 | 87.8 | 84.9 | 90.7 | Joint subset partition and labeling | 94.0 | 90.0 | 83.3 | 77.2 | 82.5 | 75.6 | 68.5 | 81.6 | Convolutional pose machines | 97.6 | 94.9 | 88.5 | 83.8 | 88.1 | 82.5 | 79.1 | 87.8 | Model of this study | 97.7 | 96.0 | 90.6 | 86.0 | 89.7 | 86.3 | 82.6 | 89.7 |
|
|