Table 3.

Baseline comparison on the dataset for joint phase and step recognition. Accuracy (ACC) is reported after 4-fold cross-validation

		Phase ACC	Step ACC	Phase-Step ACC
No TCN	ResNet	82.1 ± 2.9	65.5 ± 1.8	54.9 ± 2.6
	MT-ResNet	81.7 ± 2.3	66.6 ± 2.1	64.8 ± 2.0
	ResNetLSTM	89.1 ± 2.4	71.3 ± 2.0	68.5 ± 2.3
	MT-ResNetLSTM	88.6 ± 2.3	72.2 ± 1.8	70.7 ± 1.9
Stage I	TeCNO	89.8 ± 3.0	75.1 ± 2.1	72.3 ± 3.0
	MTMS-TCN	91.2 ± 2.5	76.1 ± 2.3	75.1 ± 2.8
Stage II	TeCNO	89.9 ± 2.8	74.8 ± 2.2	71.9 ± 2.7
	MTMS-TCN	90.9 ± 2.8	75.5 ± 2.7	75.1 ± 2.8

Bold numbers denote best performance for each metric