Skip to main content
. 2020 Sep 28;20(19):5553. doi: 10.3390/s20195553

Table 4.

Stride length prediction errors, mean error (ME), mean absolute error (MAE), and normalized absolute percentage error (NAPE) for multiple contemporary network architectures.

Models ME ± STD (cm)
Validation Set
ME ± STD (cm)
Test Set
MAE ± STD (cm)
Validation Set
MAE ± STD (cm)
Test Set
NAPE ± STD (%)
Validation Set
NAPE ± STD (%)
Test Set
Number Of
Parameters
CNN [23] 0.5 ± 4.2 −2.2 ± 9.7 3.4 ± 2.9 8.2 ± 6.2 3 ± 2.5 7.2 ± 5.5 2,079,921
CNN [17] 0.4 ± 3.7 −2.4 ± 8.7 2.9 ± 2.6 7.6 ± 6.1 2.5 ± 2.2 6.8 ± 5.5 148,529
FCN [35] −2.7 ± 3.9 −4.8 ± 9.1 8.4 ± 3.5 11.9 ± 7.1 7.3 ± 3 10.5 ± 6.3 277,121
ResNet [35] 0.5 ± 3.9 −1.9 ± 9.6 5.1 ± 3.2 9.1 ± 6.4 4.4 ± 2.8 8.1 ± 5.7 229,953
MLSTM-FCN [36] 1.0 ± 3.6 −1.2 ± 9.4 6.1 ± 3.1 9.5 ± 6.8 5.3 ± 2.7 8.3 ± 6 277,801
MALSTM-FCN [36] 1.0 ± 3.7 −0.8 ± 9.0 6.9 ± 3.2 10.3 ± 6.5 5.9 ± 2.7 9.1 ± 5.7 278,361
Hannink et al. [23] NA −0.15± 6.1 NA NA 2,079,921
Zrenner et al. [17] NA 2.5 ± 20.1 NA 15.3