Research Article
Pedestrian Motion Path Detection Method Based on Deep Learning and Foreground Detection
Table 1
Model performance under different confidence thresholds.
| Confidence threshold | Transformation of output layer | Improving the network | Accuracy | Recall | F1 score | Accuracy | Recall | F1 score |
| 0.1 | 0.87 | 0.88 | 0.87 | 0.93 | 0.93 | 0.93 | 0.2 | 0.92 | 0.87 | 0.89 | 0.96 | 0.92 | 0.94 | 0.3 | 0.94 | 0.85 | 0.89 | 0.97 | 0.91 | 0.94 | 0.4 | 0.95 | 0.84 | 0.88 | 0.98 | 0.89 | 0.93 | 0.5 | 0.96 | 0.82 | 0.89 | 0.98 | 0.86 | 0.92 | 0.6 | 0.97 | 0.81 | 0.88 | 0.99 | 0.87 | 0.92 | 0.7 | 0.98 | 0.78 | 0.86 | 0.98 | 0.84 | 0.91 | 0.8 | 0.98 | 0.73 | 0.84 | 0.99 | 0.81 | 0.89 | 0.9 | 0.99 | 0.63 | 0.77 | 0.98 | 0.73 | 0.85 | Average accuracy | 0.8713 | 0.9131 |
|
|