Table 7.

The classification results for different models with offline and on-device training are compared. We show three different models (CNN-LSTM, ResNet, and DeepConvLSTM) with both offline and on-device training for the gait analysis dataset. The best scores are emboldened row-wise per metric.

Model	Offline Training		On-Device Training
	Accuracy	F1	Accuracy	F1
CNN-LSTM	0.607 $(\pm 0.12)$	0.598 $(\pm 0.11)$	0.832 $(\pm 0.16)$	0.821 $(\pm 0.13)$
ResNet	0.580 $(\pm 0.13)$	0.576 $(\pm 0.12)$	0.863 $(\pm 0.19)$	0.847 $(\pm 0.15)$
DeepConvLSTM	0.578 $(\pm 0.12)$	0.530 $(\pm 0.14)$	0.817 $(\pm 0.13)$	0.813 $(\pm 0.12)$