Table 3.
Accuracy of each computer vision method compared with the Optotrak measures.
| Methods | M-TF -value (-value) | Var-TF -value (-value) |
|---|---|---|
| ‘As fast as possible’ condition with frequency ¡ 4 Hz | ||
| ResNet50 | 0.52 (0.48) | 1.48 (0.24) |
| EfficientNetB0 | 0.72 (0.47) | 2.60 (0.01) |
| MobileNetV2 | 0.39 (0.54) | 1.89 (0.19) |
| ‘As fast as possible’ condition with frequency ¿ 4 Hz | ||
| ResNet50 | 107.61 (0) | 59.40 (0) |
| EfficientNetB0 | 7.44 (0) | 6.57 (0) |
| MobileNetV2 | 109.67 (0) | 41.56 (0) |
| 0.5 Hz condition | ||
| ResNet50 | 0.39 (0.53) | 6.41 (0.02) |
| EfficientNetB0 | 0.84 (0.40) | 1.81 (0.08) |
| MobileNetV2 | 0.02 (0.88) | 7.63 (0.01) |
| 1 Hz condition | ||
| ResNet50 | 0.02 (0.90) | 3.40 (0.70) |
| EfficientNetB0 | 0.02 (0.99) | 0.58 (0.56) |
| MobileNetV2 | 0.05 (0.82) | 3.63 (0.06) |
| 2 Hz condition | ||
| ResNet50 | 0.02 (0.89) | 6.10 (0.02) |
| EfficientNetB0 | 0.93 (0.36) | 1.43 (0.16) |
| MobileNetV2 | 0.01 (0.92) | 4.94 (0.03) |
| 3 Hz condition | ||
| ResNet50 | 4.26 (0.04) | 3.92 (0.05) |
| EfficientNetB0 | 1.61 (0.11) | 1.79 (0.08) |
| MobileNetV2 | 2.16 (0.15) | 3.63 (0.06) |