Table 5 Validate experiments.
From: A hybrid human fall detection method based on modified YOLOv8s and AlphaPose
Paper | Method | Precision (%) | Recall (%) | Accuracy (%) | F1 | FPS (f/s) |
|---|---|---|---|---|---|---|
Koo35 | CNN+LSTM | 70.6 | 68.4 | 69.9 | 0.695 | 21.9 |
Nguyen36 | YOLOv5+HRNet | 72.5 | 71.3 | 72.1 | 0.719 | 17.3 |
Dey37 | Transformer | 76.4 | 74.8 | 75.8 | 0.756 | 12.5 |
Wei38 | Mediapipe+LSTM | 72.1 | 70.6 | 71.7 | 0.713 | 13.6 |
Gao39 | OpenPose+MobileNet | 73.2 | 71.4 | 72.6 | 0.723 | 17 |
Inturi40 | 2s-AGCN | 80.1 | 79.2 | 79.4 | 0.796 | 8.3 |
Lee41 | Alphapose+LSTM | 74.7 | 73.3 | 74.3 | 0.74 | 14.4 |
Ours | HFDMIA-Pose | 78.7 | 76.1 | 77.5 | 0.774 | 19.8 |
